diff --git a/src/gallium/state_trackers/dri/dri_drawable.c b/src/gallium/state_trackers/dri/dri_drawable.c index 0a952f7b284..acec7190374 100644 --- a/src/gallium/state_trackers/dri/dri_drawable.c +++ b/src/gallium/state_trackers/dri/dri_drawable.c @@ -44,14 +44,6 @@ #include "util/u_memory.h" -static void -dri_copy_to_front(__DRIdrawablePrivate * dPriv, - struct pipe_surface *from, - int x, int y, unsigned w, unsigned h) -{ - /* TODO send a message to the Xserver to copy to the real front buffer */ -} - static struct pipe_surface * dri_surface_from_handle(struct drm_api *api, struct pipe_screen *screen, @@ -95,6 +87,35 @@ dri_surface_from_handle(struct drm_api *api, return surface; } +/** + * Pixmaps have will have the same name of fake front and front. + */ +static boolean +dri2_check_if_pixmap(__DRIbuffer *buffers, int count) +{ + boolean found = FALSE; + boolean is_pixmap = FALSE; + unsigned name; + int i; + + for (i = 0; i < count; i++) { + switch (buffers[i].attachment) { + case __DRI_BUFFER_FRONT_LEFT: + case __DRI_BUFFER_FAKE_FRONT_LEFT: + if (found) { + is_pixmap = buffers[i].name == name; + } else { + name = buffers[i].name; + found = TRUE; + } + default: + continue; + } + } + + return is_pixmap; +} + /** * This will be called a drawable is known to have been resized. */ @@ -153,15 +174,15 @@ dri_get_buffers(__DRIdrawablePrivate * dPriv) memcpy(drawable->old, buffers, sizeof(__DRIbuffer) * count); } + drawable->is_pixmap = dri2_check_if_pixmap(buffers, count); + for (i = 0; i < count; i++) { enum pipe_format format = 0; int index = 0; switch (buffers[i].attachment) { case __DRI_BUFFER_FRONT_LEFT: - index = ST_SURFACE_FRONT_LEFT; - format = drawable->color_format; - break; + continue; case __DRI_BUFFER_FAKE_FRONT_LEFT: index = ST_SURFACE_FRONT_LEFT; format = drawable->color_format; @@ -233,8 +254,25 @@ dri_flush_frontbuffer(struct pipe_screen *screen, struct pipe_surface *surf, void *context_private) { struct dri_context *ctx = (struct dri_context *)context_private; + struct dri_drawable *drawable = dri_drawable(ctx->dPriv); + __DRIdrawable *dri_drawable = ctx->dPriv; + __DRIscreen *dri_screen = ctx->sPriv; - dri_copy_to_front(ctx->dPriv, surf, 0, 0, surf->width, surf->height); + /* XXX Does this function get called with DRI1? */ + + if (ctx->dPriv == NULL) { + debug_printf("%s: no drawable bound to context\n", __func__); + return; + } + +#if 0 + /* TODO if rendering to pixmaps is slow enable this code. */ + if (drawable->is_pixmap) + return; +#endif + + (*dri_screen->dri2.loader->flushFrontBuffer)(dri_drawable, + dri_drawable->loaderPrivate); } /** @@ -259,26 +297,20 @@ dri_create_buffer(__DRIscreenPrivate * sPriv, drawable->color_format = (visual->redBits == 8) ? PIPE_FORMAT_A8R8G8B8_UNORM : PIPE_FORMAT_R5G6B5_UNORM; - debug_printf("Red bits is %d\n", visual->redBits); - switch(visual->depthBits) { default: case 0: - debug_printf("Depth buffer 0.\n"); drawable->depth_format = PIPE_FORMAT_NONE; break; case 16: - debug_printf("Depth buffer 16.\n"); drawable->depth_format = PIPE_FORMAT_Z16_UNORM; break; case 24: if (visual->stencilBits == 0) { - debug_printf("Depth buffer 24. Stencil 0.\n"); drawable->depth_format = (screen->d_depth_bits_last) ? PIPE_FORMAT_X8Z24_UNORM: PIPE_FORMAT_Z24X8_UNORM; } else { - debug_printf("Combined depth stencil 24 / 8.\n"); drawable->depth_format = (screen->sd_depth_bits_last) ? PIPE_FORMAT_S8Z24_UNORM: PIPE_FORMAT_Z24S8_UNORM; @@ -312,12 +344,11 @@ dri_create_buffer(__DRIscreenPrivate * sPriv, dPriv->driverPrivate = (void *)drawable; /* setup dri2 buffers information */ + /* TODO incase of double buffer visual, delay fake creation */ i = 0; drawable->attachments[i++] = __DRI_BUFFER_FRONT_LEFT; -#if 0 - /* TODO incase of double buffer visual, delay fake creation */ drawable->attachments[i++] = __DRI_BUFFER_FAKE_FRONT_LEFT; -#endif + if (visual->doubleBufferMode) drawable->attachments[i++] = __DRI_BUFFER_BACK_LEFT; if (visual->depthBits) diff --git a/src/gallium/state_trackers/dri/dri_drawable.h b/src/gallium/state_trackers/dri/dri_drawable.h index dfd0b8766d2..eaf0b954bd4 100644 --- a/src/gallium/state_trackers/dri/dri_drawable.h +++ b/src/gallium/state_trackers/dri/dri_drawable.h @@ -46,6 +46,8 @@ struct dri_drawable unsigned attachments[8]; unsigned num_attachments; + boolean is_pixmap; + __DRIbuffer old[8]; unsigned old_num; unsigned old_w; diff --git a/src/gallium/state_trackers/xorg/xorg_dri2.c b/src/gallium/state_trackers/xorg/xorg_dri2.c index 3fbab4dc51d..f089965b03c 100644 --- a/src/gallium/state_trackers/xorg/xorg_dri2.c +++ b/src/gallium/state_trackers/xorg/xorg_dri2.c @@ -97,6 +97,11 @@ driCreateBuffers(DrawablePtr pDraw, unsigned int *attachments, int count) template.tex_usage = PIPE_TEXTURE_USAGE_DEPTH_STENCIL; tex = ms->screen->texture_create(ms->screen, &template); depth = tex; + } else if (attachments[i] == DRI2BufferFakeFrontLeft && + pDraw->type == DRAWABLE_PIXMAP) { + pPixmap = (PixmapPtr) pDraw; + pPixmap->refcnt++; + tex = xorg_exa_get_texture(pPixmap); } else { pPixmap = (*pScreen->CreatePixmap)(pScreen, pDraw->width, pDraw->height, @@ -171,12 +176,53 @@ driCopyRegion(DrawablePtr pDraw, RegionPtr pRegion, GCPtr gc; RegionPtr copy_clip; + /* + * In driCreateBuffers we dewrap windows into the + * backing pixmaps in order to get to the texture. + * We need to use the real drawable in CopyArea + * so that cliprects and offsets are correct. + */ src_pixmap = src_priv->pPixmap; dst_pixmap = dst_priv->pPixmap; if (pSrcBuffer->attachment == DRI2BufferFrontLeft) src_pixmap = (PixmapPtr)pDraw; if (pDestBuffer->attachment == DRI2BufferFrontLeft) dst_pixmap = (PixmapPtr)pDraw; + + /* + * The clients implements glXWaitX with a copy front to fake and then + * waiting on the server to signal its completion of it. While + * glXWaitGL is a client side flush and a copy from fake to front. + * This is how it is done in the DRI2 protocol, how ever depending + * which type of drawables the server does things a bit differently + * then what the protocol says as the fake and front are the same. + * + * for pixmaps glXWaitX is a server flush. + * for pixmaps glXWaitGL is a client flush. + * for windows glXWaitX is a copy from front to fake then a server flush. + * for windows glXWaitGL is a client flush then a copy from fake to front. + * + * XXX in the windows case this code always flushes but that isn't a + * must in the glXWaitGL case but we don't know if this is a glXWaitGL + * or a glFlush/glFinish call. + */ + if (dst_pixmap == src_pixmap) { + /* pixmap glXWaitX */ + if (pSrcBuffer->attachment == DRI2BufferFrontLeft && + pDestBuffer->attachment == DRI2BufferFakeFrontLeft) { + ms->ctx->flush(ms->ctx, PIPE_FLUSH_SWAPBUFFERS, NULL); + return; + } + /* pixmap glXWaitGL */ + if (pDestBuffer->attachment == DRI2BufferFrontLeft && + pSrcBuffer->attachment == DRI2BufferFakeFrontLeft) { + return; + } else { + xf86DrvMsg(pScrn->scrnIndex, X_WARNING, + "copying between the same pixmap\n"); + } + } + gc = GetScratchGC(pDraw->depth, pScreen); copy_clip = REGION_CREATE(pScreen, NULL, 0); REGION_COPY(pScreen, copy_clip, pRegion); diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_context.c b/src/gallium/winsys/drm/intel/gem/intel_be_context.c index 629987c6f92..ff4518f8680 100644 --- a/src/gallium/winsys/drm/intel/gem/intel_be_context.c +++ b/src/gallium/winsys/drm/intel/gem/intel_be_context.c @@ -57,9 +57,6 @@ intel_be_batch_flush(struct i915_winsys *sws, struct intel_be_context *intel = intel_be_context(sws); struct intel_be_fence **f = (struct intel_be_fence **)fence; - if (fence && *fence) - assert(0); - intel_be_batchbuffer_flush(intel->batch, f); } diff --git a/src/glx/x11/dri2_glx.c b/src/glx/x11/dri2_glx.c index d5d5a07fee3..89efe3ab292 100644 --- a/src/glx/x11/dri2_glx.c +++ b/src/glx/x11/dri2_glx.c @@ -159,7 +159,7 @@ dri2DestroyDrawable(__GLXDRIdrawable * pdraw) const __DRIcoreExtension *core = pdraw->psc->core; (*core->destroyDrawable) (pdraw->driDrawable); - DRI2DestroyDrawable(pdraw->psc->dpy, pdraw->drawable); + DRI2DestroyDrawable(pdraw->psc->dpy, pdraw->xDrawable); Xfree(pdraw); } @@ -189,7 +189,7 @@ dri2CreateDrawable(__GLXscreenConfigs * psc, config->driConfig, pdraw); if (!pdraw->base.driDrawable) { - DRI2DestroyDrawable(psc->dpy, drawable); + DRI2DestroyDrawable(psc->dpy, xDrawable); Xfree(pdraw); return NULL; } @@ -221,7 +221,7 @@ dri2CopySubBuffer(__GLXDRIdrawable * pdraw, region = XFixesCreateRegion(pdraw->psc->dpy, &xrect, 1); /* should get a fence ID back from here at some point */ - DRI2CopyRegion(pdraw->psc->dpy, pdraw->drawable, region, + DRI2CopyRegion(pdraw->psc->dpy, pdraw->xDrawable, region, DRI2BufferFrontLeft, DRI2BufferBackLeft); XFixesDestroyRegion(pdraw->psc->dpy, region); @@ -261,7 +261,7 @@ dri2WaitX(__GLXDRIdrawable * pdraw) #endif region = XFixesCreateRegion(pdraw->psc->dpy, &xrect, 1); - DRI2CopyRegion(pdraw->psc->dpy, pdraw->drawable, region, + DRI2CopyRegion(pdraw->psc->dpy, pdraw->xDrawable, region, DRI2BufferFakeFrontLeft, DRI2BufferFrontLeft); XFixesDestroyRegion(pdraw->psc->dpy, region); } @@ -287,7 +287,7 @@ dri2WaitGL(__GLXDRIdrawable * pdraw) #endif region = XFixesCreateRegion(pdraw->psc->dpy, &xrect, 1); - DRI2CopyRegion(pdraw->psc->dpy, pdraw->drawable, region, + DRI2CopyRegion(pdraw->psc->dpy, pdraw->xDrawable, region, DRI2BufferFrontLeft, DRI2BufferFakeFrontLeft); XFixesDestroyRegion(pdraw->psc->dpy, region); } diff --git a/src/glx/x11/glxcmds.c b/src/glx/x11/glxcmds.c index 7eb23dbacab..2efe1919820 100644 --- a/src/glx/x11/glxcmds.c +++ b/src/glx/x11/glxcmds.c @@ -1960,7 +1960,7 @@ __glXBeginFrameTrackingMESA(Display * dpy, GLXDrawable drawable) { int status = GLX_BAD_CONTEXT; #ifdef __DRI_FRAME_TRACKING - int screen; + int screen = 0; __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable, &screen); __GLXscreenConfigs *const psc = GetGLXScreenConfigs(dpy, screen); @@ -1979,7 +1979,7 @@ __glXEndFrameTrackingMESA(Display * dpy, GLXDrawable drawable) { int status = GLX_BAD_CONTEXT; #ifdef __DRI_FRAME_TRACKING - int screen; + int screen = 0; __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable, &screen); __GLXscreenConfigs *psc = GetGLXScreenConfigs(dpy, screen); @@ -1999,7 +1999,7 @@ __glXGetFrameUsageMESA(Display * dpy, GLXDrawable drawable, GLfloat * usage) { int status = GLX_BAD_CONTEXT; #ifdef __DRI_FRAME_TRACKING - int screen; + int screen = 0; __GLXDRIdrawable *const pdraw = GetGLXDRIDrawable(dpy, drawable, &screen); __GLXscreenConfigs *const psc = GetGLXScreenConfigs(dpy, screen); @@ -2029,7 +2029,7 @@ __glXQueryFrameTrackingMESA(Display * dpy, GLXDrawable drawable, { int status = GLX_BAD_CONTEXT; #ifdef __DRI_FRAME_TRACKING - int screen; + int screen = 0; __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable, &screen); __GLXscreenConfigs *const psc = GetGLXScreenConfigs(dpy, screen); @@ -2213,7 +2213,7 @@ PUBLIC GLXFBConfigSGIX glXGetFBConfigFromVisualSGIX(Display * dpy, XVisualInfo * vis) { __GLXdisplayPrivate *priv; - __GLXscreenConfigs *psc; + __GLXscreenConfigs *psc = NULL; if ((GetGLXPrivScreenConfig(dpy, vis->screen, &priv, &psc) != Success) && __glXExtensionBitIsEnabled(psc, SGIX_fbconfig_bit) @@ -2432,7 +2432,7 @@ __glXWaitForMscOML(Display * dpy, GLXDrawable drawable, int64_t * msc, int64_t * sbc) { #ifdef __DRI_MEDIA_STREAM_COUNTER - int screen; + int screen = 0; __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable, &screen); __GLXscreenConfigs *const psc = GetGLXScreenConfigs(dpy, screen); int ret; diff --git a/src/mesa/drivers/dri/i965/brw_wm_emit.c b/src/mesa/drivers/dri/i965/brw_wm_emit.c index b3cf524c63e..5f29ca1cd49 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_emit.c +++ b/src/mesa/drivers/dri/i965/brw_wm_emit.c @@ -588,17 +588,19 @@ static void emit_dph( struct brw_compile *p, const struct brw_reg *arg0, const struct brw_reg *arg1 ) { + const int dst_chan = _mesa_ffs(mask & WRITEMASK_XYZW) - 1; + if (!(mask & WRITEMASK_XYZW)) return; /* Do not emit dead code */ - assert((mask & WRITEMASK_XYZW) == WRITEMASK_X); + assert(is_power_of_two(mask & WRITEMASK_XYZW)); brw_MUL(p, brw_null_reg(), arg0[0], arg1[0]); brw_MAC(p, brw_null_reg(), arg0[1], arg1[1]); - brw_MAC(p, dst[0], arg0[2], arg1[2]); + brw_MAC(p, dst[dst_chan], arg0[2], arg1[2]); brw_set_saturate(p, (mask & SATURATE) ? 1 : 0); - brw_ADD(p, dst[0], dst[0], arg1[3]); + brw_ADD(p, dst[dst_chan], dst[dst_chan], arg1[3]); brw_set_saturate(p, 0); } diff --git a/src/mesa/drivers/dri/r300/r300_cmdbuf.c b/src/mesa/drivers/dri/r300/r300_cmdbuf.c index b08af116a12..87651716fef 100644 --- a/src/mesa/drivers/dri/r300/r300_cmdbuf.c +++ b/src/mesa/drivers/dri/r300/r300_cmdbuf.c @@ -105,7 +105,7 @@ void emit_vpu(GLcontext *ctx, struct radeon_state_atom * atom) ndw -= 5; OUT_BATCH_REGVAL(R300_VAP_PVS_VECTOR_INDX_REG, addr); OUT_BATCH(CP_PACKET0(R300_VAP_PVS_UPLOAD_DATA, ndw-1) | RADEON_ONE_REG_WR); - OUT_BATCH_TABLE(atom->cmd + 1, ndw); + OUT_BATCH_TABLE(&atom->cmd[1], ndw); OUT_BATCH_REGVAL(R300_VAP_PVS_STATE_FLUSH_REG, 0); END_BATCH(); } @@ -134,7 +134,7 @@ void emit_r500fp(GLcontext *ctx, struct radeon_state_atom * atom) OUT_BATCH(addr); ndw-=3; OUT_BATCH(CP_PACKET0(R500_GA_US_VECTOR_DATA, ndw-1) | RADEON_ONE_REG_WR); - OUT_BATCH_TABLE(atom->cmd + 1, ndw); + OUT_BATCH_TABLE(&atom->cmd[1], ndw); END_BATCH(); } diff --git a/src/mesa/drivers/dri/r300/r300_draw.c b/src/mesa/drivers/dri/r300/r300_draw.c index 6652b396b9b..2ac898bd59d 100644 --- a/src/mesa/drivers/dri/r300/r300_draw.c +++ b/src/mesa/drivers/dri/r300/r300_draw.c @@ -109,11 +109,11 @@ static void r300FixupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer #if MESA_BIG_ENDIAN } else { /* if (mesa_ind_buf->type == GL_UNSIGNED_SHORT) */ - GLuint size; GLushort *in = (GLushort *)src_ptr; - size = sizeof(GLushort) * ((mesa_ind_buf->count + 1) & ~1); + GLuint size = sizeof(GLushort) * ((mesa_ind_buf->count + 1) & ~1); - radeonAllocDmaRegion(&r300->radeon, &r300->ind_buf.bo, &r300->ind_buf.bo_offet, size, 4); + radeonAllocDmaRegion(&r300->radeon, &r300->ind_buf.bo, + &r300->ind_buf.bo_offset, size, 4); assert(r300->ind_buf.bo->ptr != NULL); out = (GLuint *)ADD_POINTERS(r300->ind_buf.bo->ptr, r300->ind_buf.bo_offset); diff --git a/src/mesa/drivers/dri/r300/r300_render.c b/src/mesa/drivers/dri/r300/r300_render.c index f3674f9c7b8..446cf40131e 100644 --- a/src/mesa/drivers/dri/r300/r300_render.c +++ b/src/mesa/drivers/dri/r300/r300_render.c @@ -390,7 +390,7 @@ void r300RunRenderPrimitive(GLcontext * ctx, int start, int end, int prim) GLuint first, incr, offset = 0; if (!split_prim_inplace(prim & PRIM_MODE_MASK, &first, &incr) && - num_verts > 65500) { + num_verts > 65535) { WARN_ONCE("Fixme: can't handle spliting prim %d\n", prim); return; } diff --git a/src/mesa/drivers/dri/r600/r600_cmdbuf.c b/src/mesa/drivers/dri/r600/r600_cmdbuf.c index 38814b6d71e..10ea7668963 100644 --- a/src/mesa/drivers/dri/r600/r600_cmdbuf.c +++ b/src/mesa/drivers/dri/r600/r600_cmdbuf.c @@ -322,7 +322,7 @@ static int r600_cs_emit(struct radeon_cs *cs) struct drm_radeon_cs_chunk cs_chunk[2]; uint32_t length_dw_reloc_chunk; uint64_t chunk_ptrs[2]; - uint32_t reloc_chunk[128]; + uint32_t reloc_chunk[256]; int r; int retry = 0; diff --git a/src/mesa/drivers/dri/r600/r600_context.c b/src/mesa/drivers/dri/r600/r600_context.c index 4489064c0d0..0b0c4f5049b 100644 --- a/src/mesa/drivers/dri/r600/r600_context.c +++ b/src/mesa/drivers/dri/r600/r600_context.c @@ -185,7 +185,13 @@ static void r600_vtbl_emit_cs_header(struct radeon_cs *cs, radeonContextPtr rmes static void r600_vtbl_pre_emit_atoms(radeonContextPtr radeon) { - /* to be enabled */ + context_t *context = (context_t *)radeon; + + /* always emit CB base to prevent + * lock ups on some chips. + */ + R600_STATECHANGE(context, cb_target); + r700Start3D(context); } static void r600_fallback(GLcontext *ctx, GLuint bit, GLboolean mode) diff --git a/src/mesa/drivers/dri/r600/r600_context.h b/src/mesa/drivers/dri/r600/r600_context.h index a9b080baa31..8ae05a301c7 100644 --- a/src/mesa/drivers/dri/r600/r600_context.h +++ b/src/mesa/drivers/dri/r600/r600_context.h @@ -55,10 +55,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. struct r600_context; typedef struct r600_context context_t; -extern GLboolean r700SendPSState(context_t *context); -extern GLboolean r700SendVSState(context_t *context); -extern GLboolean r700SendFSState(context_t *context); - #include "main/mm.h" /************ DMA BUFFERS **************/ @@ -115,17 +111,37 @@ enum struct r600_hw_state { struct radeon_state_atom sq; struct radeon_state_atom db; + struct radeon_state_atom stencil; struct radeon_state_atom db_target; struct radeon_state_atom sc; + struct radeon_state_atom scissor; + struct radeon_state_atom aa; struct radeon_state_atom cl; + struct radeon_state_atom gb; struct radeon_state_atom ucp; struct radeon_state_atom su; + struct radeon_state_atom poly; struct radeon_state_atom cb; + struct radeon_state_atom clrcmp; + struct radeon_state_atom blnd; + struct radeon_state_atom blnd_clr; struct radeon_state_atom cb_target; struct radeon_state_atom sx; struct radeon_state_atom vgt; struct radeon_state_atom spi; struct radeon_state_atom vpt; + + struct radeon_state_atom fs; + struct radeon_state_atom vs; + struct radeon_state_atom ps; + + struct radeon_state_atom vs_consts; + struct radeon_state_atom ps_consts; + + struct radeon_state_atom vtx; + struct radeon_state_atom tx; + struct radeon_state_atom tx_smplr; + struct radeon_state_atom tx_brdr_clr; }; /** @@ -168,22 +184,14 @@ do { \ r600->radeon.hw.is_dirty = GL_TRUE; \ } while(0) -extern GLboolean r700SendTextureState(context_t *context); - extern GLboolean r700SyncSurf(context_t *context, struct radeon_bo *pbo, uint32_t read_domain, uint32_t write_domain, uint32_t sync_type); -extern int r700SetupStreams(GLcontext * ctx); -extern void r700SetupVTXConstants(GLcontext * ctx, - unsigned int nStreamID, - void * pAos, - unsigned int size, /* number of elements in vector */ - unsigned int stride, - unsigned int Count); /* number of vectors in stream */ - +extern void r700SetupStreams(GLcontext * ctx); +extern void r700Start3D(context_t *context); extern void r600InitAtoms(context_t *context); #define RADEON_D_CAPTURE 0 diff --git a/src/mesa/drivers/dri/r600/r600_emit.c b/src/mesa/drivers/dri/r600/r600_emit.c index b0c7294682a..be86de170f9 100644 --- a/src/mesa/drivers/dri/r600/r600_emit.c +++ b/src/mesa/drivers/dri/r600/r600_emit.c @@ -51,53 +51,55 @@ void r600EmitCacheFlush(context_t *rmesa) { } -GLboolean r600EmitShader(GLcontext * ctx, +GLboolean r600EmitShader(GLcontext * ctx, void ** shaderbo, - GLvoid * data, + GLvoid * data, int sizeinDWORD, - char * szShaderUsage) + char * szShaderUsage) { - radeonContextPtr radeonctx = RADEON_CONTEXT(ctx); - - struct radeon_bo * pbo; - uint32_t *out; + radeonContextPtr radeonctx = RADEON_CONTEXT(ctx); + struct radeon_bo * pbo; + uint32_t *out; shader_again_alloc: - pbo = radeon_bo_open(radeonctx->radeonScreen->bom, - 0, - sizeinDWORD * 4, - 256, - RADEON_GEM_DOMAIN_GTT, - 0); + pbo = radeon_bo_open(radeonctx->radeonScreen->bom, + 0, + sizeinDWORD * 4, + 256, + RADEON_GEM_DOMAIN_GTT, + 0); if (!pbo) { rcommonFlushCmdBuf(radeonctx, __FUNCTION__); goto shader_again_alloc; } + radeon_cs_space_add_persistent_bo(radeonctx->cmdbuf.cs, + pbo, + RADEON_GEM_DOMAIN_GTT, 0); + if (radeon_cs_space_check_with_bo(radeonctx->cmdbuf.cs, pbo, - RADEON_GEM_DOMAIN_GTT, 0)) + RADEON_GEM_DOMAIN_GTT, 0)) { fprintf(stderr,"failure to revalidate BOs - badness\n"); - + return GL_FALSE; + } radeon_bo_map(pbo, 1); - radeon_bo_ref(pbo); + out = (uint32_t*)(pbo->ptr); - out = (uint32_t*)(pbo->ptr); + memcpy(out, data, sizeinDWORD * 4); - memcpy(out, data, sizeinDWORD * 4); + radeon_bo_unmap(pbo); - radeon_bo_unmap(pbo); + *shaderbo = (void*)pbo; - *shaderbo = (void*)pbo; - - return GL_TRUE; + return GL_TRUE; } -GLboolean r600DeleteShader(GLcontext * ctx, - void * shaderbo) +GLboolean r600DeleteShader(GLcontext * ctx, + void * shaderbo) { struct radeon_bo * pbo = (struct radeon_bo *)shaderbo; diff --git a/src/mesa/drivers/dri/r600/r600_texstate.c b/src/mesa/drivers/dri/r600/r600_texstate.c index 1057d7d8bbf..9f1bf45246f 100644 --- a/src/mesa/drivers/dri/r600/r600_texstate.c +++ b/src/mesa/drivers/dri/r600/r600_texstate.c @@ -60,6 +60,10 @@ void r600UpdateTextureState(GLcontext * ctx) struct radeon_tex_obj *t; GLuint unit; + R600_STATECHANGE(context, tx); + R600_STATECHANGE(context, tx_smplr); + R600_STATECHANGE(context, tx_brdr_clr); + for (unit = 0; unit < R700_MAX_TEXTURE_UNITS; unit++) { texUnit = &ctx->Texture.Unit[unit]; t = radeon_tex_obj(ctx->Texture.Unit[unit]._Current); diff --git a/src/mesa/drivers/dri/r600/r700_chip.c b/src/mesa/drivers/dri/r600/r700_chip.c index 2d68f021dfc..3f11cf2c981 100644 --- a/src/mesa/drivers/dri/r600/r700_chip.c +++ b/src/mesa/drivers/dri/r600/r700_chip.c @@ -41,75 +41,102 @@ #include "radeon_mipmap_tree.h" -GLboolean r700SendTextureState(context_t *context) +static void r700SendTexState(GLcontext *ctx, struct radeon_state_atom *atom) { - unsigned int i; - R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw); - struct radeon_bo *bo = NULL; - BATCH_LOCALS(&context->radeon); + context_t *context = R700_CONTEXT(ctx); + R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw); + struct radeon_bo *bo = NULL; + unsigned int i; + BATCH_LOCALS(&context->radeon); - for (i=0; itextures[i]; - if (t) { - if (!t->image_override) - bo = t->mt->bo; - else - bo = t->bo; - if (bo) { + for (i = 0; i < R700_TEXTURE_NUMBERUNITS; i++) { + radeonTexObj *t = r700->textures[i]; + if (t) { + if (!t->image_override) + bo = t->mt->bo; + else + bo = t->bo; + if (bo) { - r700SyncSurf(context, bo, - RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, - 0, TC_ACTION_ENA_bit); + r700SyncSurf(context, bo, + RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, + 0, TC_ACTION_ENA_bit); - BEGIN_BATCH_NO_AUTOSTATE(9 + 4); - R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_RESOURCE, 7)); - R600_OUT_BATCH(i * 7); - R600_OUT_BATCH(r700->textures[i]->SQ_TEX_RESOURCE0); - R600_OUT_BATCH(r700->textures[i]->SQ_TEX_RESOURCE1); - R600_OUT_BATCH(0); /* r700->textures[i]->SQ_TEX_RESOURCE2 */ - R600_OUT_BATCH(r700->textures[i]->SQ_TEX_RESOURCE3); - R600_OUT_BATCH(r700->textures[i]->SQ_TEX_RESOURCE4); - R600_OUT_BATCH(r700->textures[i]->SQ_TEX_RESOURCE5); - R600_OUT_BATCH(r700->textures[i]->SQ_TEX_RESOURCE6); - R600_OUT_BATCH_RELOC(r700->textures[i]->SQ_TEX_RESOURCE2, - bo, - 0, - RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0); - R600_OUT_BATCH_RELOC(r700->textures[i]->SQ_TEX_RESOURCE3, - bo, - r700->textures[i]->SQ_TEX_RESOURCE3, - RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0); - END_BATCH(); - - BEGIN_BATCH_NO_AUTOSTATE(5); - R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_SAMPLER, 3)); - R600_OUT_BATCH(i * 3); - R600_OUT_BATCH(r700->textures[i]->SQ_TEX_SAMPLER0); - R600_OUT_BATCH(r700->textures[i]->SQ_TEX_SAMPLER1); - R600_OUT_BATCH(r700->textures[i]->SQ_TEX_SAMPLER2); - END_BATCH(); - - BEGIN_BATCH_NO_AUTOSTATE(2 + 4); - R600_OUT_BATCH_REGSEQ((TD_PS_SAMPLER0_BORDER_RED + (i * 16)), 4); - R600_OUT_BATCH(r700->textures[i]->TD_PS_SAMPLER0_BORDER_RED); - R600_OUT_BATCH(r700->textures[i]->TD_PS_SAMPLER0_BORDER_GREEN); - R600_OUT_BATCH(r700->textures[i]->TD_PS_SAMPLER0_BORDER_BLUE); - R600_OUT_BATCH(r700->textures[i]->TD_PS_SAMPLER0_BORDER_ALPHA); - END_BATCH(); - - COMMIT_BATCH(); - } - } - } - return GL_TRUE; + BEGIN_BATCH_NO_AUTOSTATE(9 + 4); + R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_RESOURCE, 7)); + R600_OUT_BATCH(i * 7); + R600_OUT_BATCH(r700->textures[i]->SQ_TEX_RESOURCE0); + R600_OUT_BATCH(r700->textures[i]->SQ_TEX_RESOURCE1); + R600_OUT_BATCH(r700->textures[i]->SQ_TEX_RESOURCE2); + R600_OUT_BATCH(r700->textures[i]->SQ_TEX_RESOURCE3); + R600_OUT_BATCH(r700->textures[i]->SQ_TEX_RESOURCE4); + R600_OUT_BATCH(r700->textures[i]->SQ_TEX_RESOURCE5); + R600_OUT_BATCH(r700->textures[i]->SQ_TEX_RESOURCE6); + R600_OUT_BATCH_RELOC(r700->textures[i]->SQ_TEX_RESOURCE2, + bo, + 0, + RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0); + R600_OUT_BATCH_RELOC(r700->textures[i]->SQ_TEX_RESOURCE3, + bo, + r700->textures[i]->SQ_TEX_RESOURCE3, + RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0); + END_BATCH(); + COMMIT_BATCH(); + } + } + } } -void r700SetupVTXConstants(GLcontext * ctx, - unsigned int nStreamID, - void * pAos, - unsigned int size, /* number of elements in vector */ - unsigned int stride, - unsigned int count) /* number of vectors in stream */ +static void r700SendTexSamplerState(GLcontext *ctx, struct radeon_state_atom *atom) +{ + context_t *context = R700_CONTEXT(ctx); + R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw); + unsigned int i; + BATCH_LOCALS(&context->radeon); + + for (i = 0; i < R700_TEXTURE_NUMBERUNITS; i++) { + radeonTexObj *t = r700->textures[i]; + if (t) { + BEGIN_BATCH_NO_AUTOSTATE(5); + R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_SAMPLER, 3)); + R600_OUT_BATCH(i * 3); + R600_OUT_BATCH(r700->textures[i]->SQ_TEX_SAMPLER0); + R600_OUT_BATCH(r700->textures[i]->SQ_TEX_SAMPLER1); + R600_OUT_BATCH(r700->textures[i]->SQ_TEX_SAMPLER2); + END_BATCH(); + COMMIT_BATCH(); + } + } +} + +static void r700SendTexBorderColorState(GLcontext *ctx, struct radeon_state_atom *atom) +{ + context_t *context = R700_CONTEXT(ctx); + R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw); + unsigned int i; + BATCH_LOCALS(&context->radeon); + + for (i = 0; i < R700_TEXTURE_NUMBERUNITS; i++) { + radeonTexObj *t = r700->textures[i]; + if (t) { + BEGIN_BATCH_NO_AUTOSTATE(2 + 4); + R600_OUT_BATCH_REGSEQ((TD_PS_SAMPLER0_BORDER_RED + (i * 16)), 4); + R600_OUT_BATCH(r700->textures[i]->TD_PS_SAMPLER0_BORDER_RED); + R600_OUT_BATCH(r700->textures[i]->TD_PS_SAMPLER0_BORDER_GREEN); + R600_OUT_BATCH(r700->textures[i]->TD_PS_SAMPLER0_BORDER_BLUE); + R600_OUT_BATCH(r700->textures[i]->TD_PS_SAMPLER0_BORDER_ALPHA); + END_BATCH(); + COMMIT_BATCH(); + } + } +} + +static void r700SetupVTXConstants(GLcontext * ctx, + unsigned int nStreamID, + void * pAos, + unsigned int size, /* number of elements in vector */ + unsigned int stride, + unsigned int count) /* number of vectors in stream */ { context_t *context = R700_CONTEXT(ctx); struct radeon_aos * paos = (struct radeon_aos *)pAos; @@ -127,6 +154,7 @@ void r700SetupVTXConstants(GLcontext * ctx, if ((context->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV610) || (context->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV620) || (context->radeon.radeonScreen->chip_family == CHIP_FAMILY_RS780) || + (context->radeon.radeonScreen->chip_family == CHIP_FAMILY_RS880) || (context->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV710)) r700SyncSurf(context, paos->bo, RADEON_GEM_DOMAIN_GTT, 0, TC_ACTION_ENA_bit); else @@ -169,19 +197,38 @@ void r700SetupVTXConstants(GLcontext * ctx, } -int r700SetupStreams(GLcontext * ctx) +void r700SetupStreams(GLcontext *ctx) { context_t *context = R700_CONTEXT(ctx); - BATCH_LOCALS(&context->radeon); + struct r700_vertex_program *vpc + = (struct r700_vertex_program *)ctx->VertexProgram._Current; + TNLcontext *tnl = TNL_CONTEXT(ctx); + struct vertex_buffer *vb = &tnl->vb; + unsigned int i, j = 0; + R600_STATECHANGE(context, vtx); + + for(i=0; imesa_program.Base.InputsRead & (1 << i)) { + rcommon_emit_vector(ctx, + &context->radeon.tcl.aos[j], + vb->AttribPtr[i]->data, + vb->AttribPtr[i]->size, + vb->AttribPtr[i]->stride, + vb->Count); + j++; + } + } + context->radeon.tcl.aos_count = j; +} + +static void r700SendVTXState(GLcontext *ctx, struct radeon_state_atom *atom) +{ + context_t *context = R700_CONTEXT(ctx); struct r700_vertex_program *vpc = (struct r700_vertex_program *)ctx->VertexProgram._Current; - - TNLcontext *tnl = TNL_CONTEXT(ctx); - struct vertex_buffer *vb = &tnl->vb; - - unsigned int unBit; unsigned int i, j = 0; + BATCH_LOCALS(&context->radeon); BEGIN_BATCH_NO_AUTOSTATE(6); R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CTL_CONST, 1)); @@ -194,31 +241,18 @@ int r700SetupStreams(GLcontext * ctx) END_BATCH(); COMMIT_BATCH(); - for(i=0; imesa_program.Base.InputsRead & unBit) - { - rcommon_emit_vector(ctx, - &context->radeon.tcl.aos[j], - vb->AttribPtr[i]->data, - vb->AttribPtr[i]->size, - vb->AttribPtr[i]->stride, - vb->Count); - - /* currently aos are packed */ - r700SetupVTXConstants(ctx, - i, - (void*)(&context->radeon.tcl.aos[j]), - (unsigned int)context->radeon.tcl.aos[j].components, - (unsigned int)context->radeon.tcl.aos[j].stride * 4, - (unsigned int)context->radeon.tcl.aos[j].count); - j++; - } - } - context->radeon.tcl.aos_count = j; - - return R600_FALLBACK_NONE; + for(i=0; imesa_program.Base.InputsRead & (1 << i)) { + /* currently aos are packed */ + r700SetupVTXConstants(ctx, + i, + (void*)(&context->radeon.tcl.aos[j]), + (unsigned int)context->radeon.tcl.aos[j].components, + (unsigned int)context->radeon.tcl.aos[j].stride * 4, + (unsigned int)context->radeon.tcl.aos[j].count); + j++; + } + } } static void r700SendDepthTargetState(GLcontext *ctx, struct radeon_state_atom *atom) @@ -309,8 +343,9 @@ static void r700SendRenderTargetState(GLcontext *ctx, struct radeon_state_atom * } -GLboolean r700SendPSState(context_t *context) +static void r700SendPSState(GLcontext *ctx, struct radeon_state_atom *atom) { + context_t *context = R700_CONTEXT(ctx); R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); struct radeon_bo * pbo; BATCH_LOCALS(&context->radeon); @@ -318,7 +353,7 @@ GLboolean r700SendPSState(context_t *context) pbo = (struct radeon_bo *)r700GetActiveFpShaderBo(GL_CONTEXT(context)); if (!pbo) - return GL_FALSE; + return; r700SyncSurf(context, pbo, RADEON_GEM_DOMAIN_GTT, 0, SH_ACTION_ENA_bit); @@ -339,13 +374,11 @@ GLboolean r700SendPSState(context_t *context) COMMIT_BATCH(); - r700->ps.dirty = GL_FALSE; - - return GL_TRUE; } -GLboolean r700SendVSState(context_t *context) +static void r700SendVSState(GLcontext *ctx, struct radeon_state_atom *atom) { + context_t *context = R700_CONTEXT(ctx); R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); struct radeon_bo * pbo; BATCH_LOCALS(&context->radeon); @@ -353,7 +386,7 @@ GLboolean r700SendVSState(context_t *context) pbo = (struct radeon_bo *)r700GetActiveVpShaderBo(GL_CONTEXT(context)); if (!pbo) - return GL_FALSE; + return; r700SyncSurf(context, pbo, RADEON_GEM_DOMAIN_GTT, 0, SH_ACTION_ENA_bit); @@ -372,14 +405,11 @@ GLboolean r700SendVSState(context_t *context) END_BATCH(); COMMIT_BATCH(); - - r700->vs.dirty = GL_FALSE; - - return GL_TRUE; } -GLboolean r700SendFSState(context_t *context) +static void r700SendFSState(GLcontext *ctx, struct radeon_state_atom *atom) { + context_t *context = R700_CONTEXT(ctx); R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); struct radeon_bo * pbo; BATCH_LOCALS(&context->radeon); @@ -396,7 +426,7 @@ GLboolean r700SendFSState(context_t *context) /* XXX */ if (!pbo) - return GL_FALSE; + return; r700SyncSurf(context, pbo, RADEON_GEM_DOMAIN_GTT, 0, SH_ACTION_ENA_bit); @@ -416,9 +446,6 @@ GLboolean r700SendFSState(context_t *context) COMMIT_BATCH(); - r700->fs.dirty = GL_FALSE; - - return GL_TRUE; } static void r700SendViewportState(GLcontext *ctx, struct radeon_state_atom *atom) @@ -650,17 +677,13 @@ static void r700SendDBState(GLcontext *ctx, struct radeon_state_atom *atom) R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); BATCH_LOCALS(&context->radeon); - BEGIN_BATCH_NO_AUTOSTATE(27); + BEGIN_BATCH_NO_AUTOSTATE(23); R600_OUT_BATCH_REGVAL(DB_HTILE_DATA_BASE, r700->DB_HTILE_DATA_BASE.u32All); R600_OUT_BATCH_REGSEQ(DB_STENCIL_CLEAR, 2); R600_OUT_BATCH(r700->DB_STENCIL_CLEAR.u32All); R600_OUT_BATCH(r700->DB_DEPTH_CLEAR.u32All); - R600_OUT_BATCH_REGSEQ(DB_STENCILREFMASK, 2); - R600_OUT_BATCH(r700->DB_STENCILREFMASK.u32All); - R600_OUT_BATCH(r700->DB_STENCILREFMASK_BF.u32All); - R600_OUT_BATCH_REGVAL(DB_DEPTH_CONTROL, r700->DB_DEPTH_CONTROL.u32All); R600_OUT_BATCH_REGVAL(DB_SHADER_CONTROL, r700->DB_SHADER_CONTROL.u32All); @@ -675,15 +698,28 @@ static void r700SendDBState(GLcontext *ctx, struct radeon_state_atom *atom) COMMIT_BATCH(); } +static void r700SendStencilState(GLcontext *ctx, struct radeon_state_atom *atom) +{ + context_t *context = R700_CONTEXT(ctx); + R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); + BATCH_LOCALS(&context->radeon); + + BEGIN_BATCH_NO_AUTOSTATE(4); + R600_OUT_BATCH_REGSEQ(DB_STENCILREFMASK, 2); + R600_OUT_BATCH(r700->DB_STENCILREFMASK.u32All); + R600_OUT_BATCH(r700->DB_STENCILREFMASK_BF.u32All); + END_BATCH(); + COMMIT_BATCH(); +} + static void r700SendCBState(GLcontext *ctx, struct radeon_state_atom *atom) { context_t *context = R700_CONTEXT(ctx); R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); BATCH_LOCALS(&context->radeon); - unsigned int ui; if (context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770) { - BEGIN_BATCH_NO_AUTOSTATE(14); + BEGIN_BATCH_NO_AUTOSTATE(11); R600_OUT_BATCH_REGSEQ(CB_CLEAR_RED, 4); R600_OUT_BATCH(r700->CB_CLEAR_RED_R6XX.u32All); R600_OUT_BATCH(r700->CB_CLEAR_GREEN_R6XX.u32All); @@ -693,31 +729,50 @@ static void r700SendCBState(GLcontext *ctx, struct radeon_state_atom *atom) R600_OUT_BATCH(r700->CB_FOG_RED_R6XX.u32All); R600_OUT_BATCH(r700->CB_FOG_GREEN_R6XX.u32All); R600_OUT_BATCH(r700->CB_FOG_BLUE_R6XX.u32All); - /* R600 does not have per-MRT blend */ - R600_OUT_BATCH_REGVAL(CB_BLEND_CONTROL, r700->CB_BLEND_CONTROL.u32All); END_BATCH(); } - BEGIN_BATCH_NO_AUTOSTATE(22); + BEGIN_BATCH_NO_AUTOSTATE(7); R600_OUT_BATCH_REGSEQ(CB_TARGET_MASK, 2); R600_OUT_BATCH(r700->CB_TARGET_MASK.u32All); R600_OUT_BATCH(r700->CB_SHADER_MASK.u32All); - - R600_OUT_BATCH_REGSEQ(CB_BLEND_RED, 4); - R600_OUT_BATCH(r700->CB_BLEND_RED.u32All); - R600_OUT_BATCH(r700->CB_BLEND_GREEN.u32All); - R600_OUT_BATCH(r700->CB_BLEND_BLUE.u32All); - R600_OUT_BATCH(r700->CB_BLEND_ALPHA.u32All); - R600_OUT_BATCH_REGVAL(R7xx_CB_SHADER_CONTROL, r700->CB_SHADER_CONTROL.u32All); - R600_OUT_BATCH_REGVAL(CB_COLOR_CONTROL, r700->CB_COLOR_CONTROL.u32All); + END_BATCH(); + COMMIT_BATCH(); +} +static void r700SendCBCLRCMPState(GLcontext *ctx, struct radeon_state_atom *atom) +{ + context_t *context = R700_CONTEXT(ctx); + R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); + BATCH_LOCALS(&context->radeon); + + BEGIN_BATCH_NO_AUTOSTATE(6); R600_OUT_BATCH_REGSEQ(CB_CLRCMP_CONTROL, 4); R600_OUT_BATCH(r700->CB_CLRCMP_CONTROL.u32All); R600_OUT_BATCH(r700->CB_CLRCMP_SRC.u32All); R600_OUT_BATCH(r700->CB_CLRCMP_DST.u32All); R600_OUT_BATCH(r700->CB_CLRCMP_MSK.u32All); END_BATCH(); + COMMIT_BATCH(); +} + +static void r700SendCBBlendState(GLcontext *ctx, struct radeon_state_atom *atom) +{ + context_t *context = R700_CONTEXT(ctx); + R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); + BATCH_LOCALS(&context->radeon); + unsigned int ui; + + if (context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770) { + BEGIN_BATCH_NO_AUTOSTATE(3); + R600_OUT_BATCH_REGVAL(CB_BLEND_CONTROL, r700->CB_BLEND_CONTROL.u32All); + END_BATCH(); + } + + BEGIN_BATCH_NO_AUTOSTATE(3); + R600_OUT_BATCH_REGVAL(CB_COLOR_CONTROL, r700->CB_COLOR_CONTROL.u32All); + END_BATCH(); if (context->radeon.radeonScreen->chip_family > CHIP_FAMILY_R600) { for (ui = 0; ui < R700_MAX_RENDER_TARGETS; ui++) { @@ -731,7 +786,22 @@ static void r700SendCBState(GLcontext *ctx, struct radeon_state_atom *atom) } COMMIT_BATCH(); +} +static void r700SendCBBlendColorState(GLcontext *ctx, struct radeon_state_atom *atom) +{ + context_t *context = R700_CONTEXT(ctx); + R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); + BATCH_LOCALS(&context->radeon); + + BEGIN_BATCH_NO_AUTOSTATE(6); + R600_OUT_BATCH_REGSEQ(CB_BLEND_RED, 4); + R600_OUT_BATCH(r700->CB_BLEND_RED.u32All); + R600_OUT_BATCH(r700->CB_BLEND_GREEN.u32All); + R600_OUT_BATCH(r700->CB_BLEND_BLUE.u32All); + R600_OUT_BATCH(r700->CB_BLEND_ALPHA.u32All); + END_BATCH(); + COMMIT_BATCH(); } static void r700SendSUState(GLcontext *ctx, struct radeon_state_atom *atom) @@ -740,25 +810,33 @@ static void r700SendSUState(GLcontext *ctx, struct radeon_state_atom *atom) R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); BATCH_LOCALS(&context->radeon); - BEGIN_BATCH_NO_AUTOSTATE(19); + BEGIN_BATCH_NO_AUTOSTATE(9); R600_OUT_BATCH_REGVAL(PA_SU_SC_MODE_CNTL, r700->PA_SU_SC_MODE_CNTL.u32All); - R600_OUT_BATCH_REGSEQ(PA_SU_POINT_SIZE, 4); R600_OUT_BATCH(r700->PA_SU_POINT_SIZE.u32All); R600_OUT_BATCH(r700->PA_SU_POINT_MINMAX.u32All); R600_OUT_BATCH(r700->PA_SU_LINE_CNTL.u32All); R600_OUT_BATCH(r700->PA_SU_VTX_CNTL.u32All); + END_BATCH(); + COMMIT_BATCH(); +} + +static void r700SendPolyState(GLcontext *ctx, struct radeon_state_atom *atom) +{ + context_t *context = R700_CONTEXT(ctx); + R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); + BATCH_LOCALS(&context->radeon); + + BEGIN_BATCH_NO_AUTOSTATE(10); R600_OUT_BATCH_REGSEQ(PA_SU_POLY_OFFSET_DB_FMT_CNTL, 2); R600_OUT_BATCH(r700->PA_SU_POLY_OFFSET_DB_FMT_CNTL.u32All); R600_OUT_BATCH(r700->PA_SU_POLY_OFFSET_CLAMP.u32All); - R600_OUT_BATCH_REGSEQ(PA_SU_POLY_OFFSET_FRONT_SCALE, 4); R600_OUT_BATCH(r700->PA_SU_POLY_OFFSET_FRONT_SCALE.u32All); R600_OUT_BATCH(r700->PA_SU_POLY_OFFSET_FRONT_OFFSET.u32All); R600_OUT_BATCH(r700->PA_SU_POLY_OFFSET_BACK_SCALE.u32All); R600_OUT_BATCH(r700->PA_SU_POLY_OFFSET_BACK_OFFSET.u32All); - END_BATCH(); COMMIT_BATCH(); @@ -770,35 +848,43 @@ static void r700SendCLState(GLcontext *ctx, struct radeon_state_atom *atom) R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); BATCH_LOCALS(&context->radeon); - BEGIN_BATCH_NO_AUTOSTATE(18); + BEGIN_BATCH_NO_AUTOSTATE(12); R600_OUT_BATCH_REGVAL(PA_CL_CLIP_CNTL, r700->PA_CL_CLIP_CNTL.u32All); R600_OUT_BATCH_REGVAL(PA_CL_VTE_CNTL, r700->PA_CL_VTE_CNTL.u32All); R600_OUT_BATCH_REGVAL(PA_CL_VS_OUT_CNTL, r700->PA_CL_VS_OUT_CNTL.u32All); R600_OUT_BATCH_REGVAL(PA_CL_NANINF_CNTL, r700->PA_CL_NANINF_CNTL.u32All); - - R600_OUT_BATCH_REGSEQ(PA_CL_GB_VERT_CLIP_ADJ, 4); - R600_OUT_BATCH(r700->PA_CL_GB_VERT_CLIP_ADJ.u32All); - R600_OUT_BATCH(r700->PA_CL_GB_VERT_DISC_ADJ.u32All); - R600_OUT_BATCH(r700->PA_CL_GB_HORZ_CLIP_ADJ.u32All); - R600_OUT_BATCH(r700->PA_CL_GB_HORZ_DISC_ADJ.u32All); - END_BATCH(); COMMIT_BATCH(); } -// XXX need to split this up -static void r700SendSCState(GLcontext *ctx, struct radeon_state_atom *atom) +static void r700SendGBState(GLcontext *ctx, struct radeon_state_atom *atom) { context_t *context = R700_CONTEXT(ctx); R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); BATCH_LOCALS(&context->radeon); - BEGIN_BATCH_NO_AUTOSTATE(47); + BEGIN_BATCH_NO_AUTOSTATE(6); + R600_OUT_BATCH_REGSEQ(PA_CL_GB_VERT_CLIP_ADJ, 4); + R600_OUT_BATCH(r700->PA_CL_GB_VERT_CLIP_ADJ.u32All); + R600_OUT_BATCH(r700->PA_CL_GB_VERT_DISC_ADJ.u32All); + R600_OUT_BATCH(r700->PA_CL_GB_HORZ_CLIP_ADJ.u32All); + R600_OUT_BATCH(r700->PA_CL_GB_HORZ_DISC_ADJ.u32All); + END_BATCH(); + COMMIT_BATCH(); +} + +static void r700SendScissorState(GLcontext *ctx, struct radeon_state_atom *atom) +{ + context_t *context = R700_CONTEXT(ctx); + R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); + BATCH_LOCALS(&context->radeon); + + BEGIN_BATCH_NO_AUTOSTATE(22); R600_OUT_BATCH_REGSEQ(PA_SC_SCREEN_SCISSOR_TL, 2); R600_OUT_BATCH(r700->PA_SC_SCREEN_SCISSOR_TL.u32All); R600_OUT_BATCH(r700->PA_SC_SCREEN_SCISSOR_BR.u32All); - R600_OUT_BATCH_REGSEQ(PA_SC_WINDOW_OFFSET, 13); + R600_OUT_BATCH_REGSEQ(PA_SC_WINDOW_OFFSET, 12); R600_OUT_BATCH(r700->PA_SC_WINDOW_OFFSET.u32All); R600_OUT_BATCH(r700->PA_SC_WINDOW_SCISSOR_TL.u32All); R600_OUT_BATCH(r700->PA_SC_WINDOW_SCISSOR_BR.u32All); @@ -811,21 +897,89 @@ static void r700SendSCState(GLcontext *ctx, struct radeon_state_atom *atom) R600_OUT_BATCH(r700->PA_SC_CLIPRECT_2_BR.u32All); R600_OUT_BATCH(r700->PA_SC_CLIPRECT_3_TL.u32All); R600_OUT_BATCH(r700->PA_SC_CLIPRECT_3_BR.u32All); - R600_OUT_BATCH(r700->PA_SC_EDGERULE.u32All); R600_OUT_BATCH_REGSEQ(PA_SC_GENERIC_SCISSOR_TL, 2); R600_OUT_BATCH(r700->PA_SC_GENERIC_SCISSOR_TL.u32All); R600_OUT_BATCH(r700->PA_SC_GENERIC_SCISSOR_BR.u32All); + END_BATCH(); + COMMIT_BATCH(); +} +static void r700SendSCState(GLcontext *ctx, struct radeon_state_atom *atom) +{ + context_t *context = R700_CONTEXT(ctx); + R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); + BATCH_LOCALS(&context->radeon); + + BEGIN_BATCH_NO_AUTOSTATE(15); + R600_OUT_BATCH_REGVAL(R7xx_PA_SC_EDGERULE, r700->PA_SC_EDGERULE.u32All); R600_OUT_BATCH_REGVAL(PA_SC_LINE_STIPPLE, r700->PA_SC_LINE_STIPPLE.u32All); R600_OUT_BATCH_REGVAL(PA_SC_MPASS_PS_CNTL, r700->PA_SC_MPASS_PS_CNTL.u32All); R600_OUT_BATCH_REGVAL(PA_SC_MODE_CNTL, r700->PA_SC_MODE_CNTL.u32All); R600_OUT_BATCH_REGVAL(PA_SC_LINE_CNTL, r700->PA_SC_LINE_CNTL.u32All); + END_BATCH(); + COMMIT_BATCH(); +} + +static void r700SendAAState(GLcontext *ctx, struct radeon_state_atom *atom) +{ + context_t *context = R700_CONTEXT(ctx); + R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); + BATCH_LOCALS(&context->radeon); + + BEGIN_BATCH_NO_AUTOSTATE(12); R600_OUT_BATCH_REGVAL(PA_SC_AA_CONFIG, r700->PA_SC_AA_CONFIG.u32All); R600_OUT_BATCH_REGVAL(PA_SC_AA_SAMPLE_LOCS_MCTX, r700->PA_SC_AA_SAMPLE_LOCS_MCTX.u32All); R600_OUT_BATCH_REGVAL(PA_SC_AA_SAMPLE_LOCS_8S_WD1_MCTX, r700->PA_SC_AA_SAMPLE_LOCS_8S_WD1_MCTX.u32All); R600_OUT_BATCH_REGVAL(PA_SC_AA_MASK, r700->PA_SC_AA_MASK.u32All); + END_BATCH(); + COMMIT_BATCH(); +} +static void r700SendPSConsts(GLcontext *ctx, struct radeon_state_atom *atom) +{ + context_t *context = R700_CONTEXT(ctx); + R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); + int i; + BATCH_LOCALS(&context->radeon); + + if (r700->ps.num_consts == 0) + return; + + BEGIN_BATCH_NO_AUTOSTATE(2 + (r700->ps.num_consts * 4)); + R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_ALU_CONST, (r700->ps.num_consts * 4))); + /* assembler map const from very beginning. */ + R600_OUT_BATCH(SQ_ALU_CONSTANT_PS_OFFSET * 4); + for (i = 0; i < r700->ps.num_consts; i++) { + R600_OUT_BATCH(r700->ps.consts[i][0].u32All); + R600_OUT_BATCH(r700->ps.consts[i][1].u32All); + R600_OUT_BATCH(r700->ps.consts[i][2].u32All); + R600_OUT_BATCH(r700->ps.consts[i][3].u32All); + } + END_BATCH(); + COMMIT_BATCH(); +} + +static void r700SendVSConsts(GLcontext *ctx, struct radeon_state_atom *atom) +{ + context_t *context = R700_CONTEXT(ctx); + R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); + int i; + BATCH_LOCALS(&context->radeon); + + if (r700->vs.num_consts == 0) + return; + + BEGIN_BATCH_NO_AUTOSTATE(2 + (r700->vs.num_consts * 4)); + R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_ALU_CONST, (r700->vs.num_consts * 4))); + /* assembler map const from very beginning. */ + R600_OUT_BATCH(SQ_ALU_CONSTANT_VS_OFFSET * 4); + for (i = 0; i < r700->vs.num_consts; i++) { + R600_OUT_BATCH(r700->vs.consts[i][0].u32All); + R600_OUT_BATCH(r700->vs.consts[i][1].u32All); + R600_OUT_BATCH(r700->vs.consts[i][2].u32All); + R600_OUT_BATCH(r700->vs.consts[i][3].u32All); + } END_BATCH(); COMMIT_BATCH(); } @@ -835,39 +989,94 @@ static int check_always(GLcontext *ctx, struct radeon_state_atom *atom) return atom->cmd_size; } -#define ALLOC_STATE( ATOM, SZ, EMIT ) \ +static int check_vtx(GLcontext *ctx, struct radeon_state_atom *atom) +{ + context_t *context = R700_CONTEXT(ctx); + + return context->radeon.tcl.aos_count * 18; +} + +static int check_tx(GLcontext *ctx, struct radeon_state_atom *atom) +{ + context_t *context = R700_CONTEXT(ctx); + unsigned int i, count = 0; + R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw); + + for (i = 0; i < R700_TEXTURE_NUMBERUNITS; i++) { + radeonTexObj *t = r700->textures[i]; + if (t) + count++; + } + return count * 31; +} + +static int check_ps_consts(GLcontext *ctx, struct radeon_state_atom *atom) +{ + context_t *context = R700_CONTEXT(ctx); + R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw); + + return 2 + (r700->ps.num_consts * 4); +} + +static int check_vs_consts(GLcontext *ctx, struct radeon_state_atom *atom) +{ + context_t *context = R700_CONTEXT(ctx); + R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw); + + return 2 + (r700->vs.num_consts * 4); +} + +#define ALLOC_STATE( ATOM, CHK, SZ, EMIT ) \ do { \ - context->atoms.ATOM.cmd_size = (SZ); \ - context->atoms.ATOM.cmd = NULL; \ - context->atoms.ATOM.name = #ATOM; \ - context->atoms.ATOM.idx = 0; \ - context->atoms.ATOM.check = check_always; \ - context->atoms.ATOM.dirty = GL_FALSE; \ - context->atoms.ATOM.emit = (EMIT); \ - context->radeon.hw.max_state_size += (SZ); \ - insert_at_tail(&context->radeon.hw.atomlist, &context->atoms.ATOM); \ + context->atoms.ATOM.cmd_size = (SZ); \ + context->atoms.ATOM.cmd = NULL; \ + context->atoms.ATOM.name = #ATOM; \ + context->atoms.ATOM.idx = 0; \ + context->atoms.ATOM.check = check_##CHK; \ + context->atoms.ATOM.dirty = GL_FALSE; \ + context->atoms.ATOM.emit = (EMIT); \ + context->radeon.hw.max_state_size += (SZ); \ + insert_at_tail(&context->radeon.hw.atomlist, &context->atoms.ATOM); \ } while (0) void r600InitAtoms(context_t *context) { + context->radeon.hw.max_state_size = 10 + 5 + 14; /* start 3d, idle, cb/db flush */ /* Setup the atom linked list */ make_empty_list(&context->radeon.hw.atomlist); context->radeon.hw.atomlist.name = "atom-list"; - ALLOC_STATE(sq, 34, r700SendSQConfig); - ALLOC_STATE(db, 27, r700SendDBState); - ALLOC_STATE(db_target, 19, r700SendDepthTargetState); - ALLOC_STATE(sc, 47, r700SendSCState); - ALLOC_STATE(cl, 18, r700SendCLState); - ALLOC_STATE(ucp, 36, r700SendUCPState); - ALLOC_STATE(su, 19, r700SendSUState); - ALLOC_STATE(cb, 39, r700SendCBState); - ALLOC_STATE(cb_target, 32, r700SendRenderTargetState); - ALLOC_STATE(sx, 9, r700SendSXState); - ALLOC_STATE(vgt, 41, r700SendVGTState); - ALLOC_STATE(spi, (59 + R700_MAX_SHADER_EXPORTS), r700SendSPIState); - ALLOC_STATE(vpt, 16, r700SendViewportState); + ALLOC_STATE(sq, always, 34, r700SendSQConfig); + ALLOC_STATE(db, always, 23, r700SendDBState); + ALLOC_STATE(stencil, always, 4, r700SendStencilState); + ALLOC_STATE(db_target, always, 12, r700SendDepthTargetState); + ALLOC_STATE(sc, always, 15, r700SendSCState); + ALLOC_STATE(scissor, always, 22, r700SendScissorState); + ALLOC_STATE(aa, always, 12, r700SendAAState); + ALLOC_STATE(cl, always, 12, r700SendCLState); + ALLOC_STATE(gb, always, 6, r700SendGBState); + ALLOC_STATE(ucp, always, 36, r700SendUCPState); + ALLOC_STATE(su, always, 9, r700SendSUState); + ALLOC_STATE(poly, always, 10, r700SendPolyState); + ALLOC_STATE(cb, always, 18, r700SendCBState); + ALLOC_STATE(clrcmp, always, 6, r700SendCBCLRCMPState); + ALLOC_STATE(blnd, always, 30, r700SendCBBlendState); + ALLOC_STATE(blnd_clr, always, 6, r700SendCBBlendColorState); + ALLOC_STATE(cb_target, always, 25, r700SendRenderTargetState); + ALLOC_STATE(sx, always, 9, r700SendSXState); + ALLOC_STATE(vgt, always, 41, r700SendVGTState); + ALLOC_STATE(spi, always, (59 + R700_MAX_SHADER_EXPORTS), r700SendSPIState); + ALLOC_STATE(vpt, always, 16, r700SendViewportState); + ALLOC_STATE(fs, always, 18, r700SendFSState); + ALLOC_STATE(vs, always, 18, r700SendVSState); + ALLOC_STATE(ps, always, 21, r700SendPSState); + ALLOC_STATE(vs_consts, vs_consts, (2 + (R700_MAX_DX9_CONSTS * 4)), r700SendVSConsts); + ALLOC_STATE(ps_consts, ps_consts, (2 + (R700_MAX_DX9_CONSTS * 4)), r700SendPSConsts); + ALLOC_STATE(vtx, vtx, (VERT_ATTRIB_MAX * 18), r700SendVTXState); + ALLOC_STATE(tx, tx, (R700_TEXTURE_NUMBERUNITS * 20), r700SendTexState); + ALLOC_STATE(tx_smplr, tx, (R700_TEXTURE_NUMBERUNITS * 5), r700SendTexSamplerState); + ALLOC_STATE(tx_brdr_clr, tx, (R700_TEXTURE_NUMBERUNITS * 6), r700SendTexBorderColorState); context->radeon.hw.is_dirty = GL_TRUE; context->radeon.hw.all_dirty = GL_TRUE; diff --git a/src/mesa/drivers/dri/r600/r700_chip.h b/src/mesa/drivers/dri/r600/r700_chip.h index c0ec4b0dd59..ae249e15fd4 100644 --- a/src/mesa/drivers/dri/r600/r700_chip.h +++ b/src/mesa/drivers/dri/r600/r700_chip.h @@ -46,6 +46,7 @@ #define R700_MAX_VIEWPORTS 16 #define R700_MAX_SHADER_EXPORTS 32 #define R700_MAX_UCP 6 +#define R700_MAX_DX9_CONSTS 256 /* Enum not show in r600_*.h */ @@ -224,6 +225,8 @@ typedef struct _PS_STATE_STRUCT union UINT_FLOAT SQ_PGM_EXPORTS_PS ; /* 0xA215 */ union UINT_FLOAT SQ_PGM_CF_OFFSET_PS ; /* 0xA233 */ GLboolean dirty; + int num_consts; + union UINT_FLOAT consts[R700_MAX_DX9_CONSTS][4]; } PS_STATE_STRUCT; typedef struct _VS_STATE_STRUCT @@ -232,6 +235,8 @@ typedef struct _VS_STATE_STRUCT union UINT_FLOAT SQ_PGM_RESOURCES_VS ; /* 0xA21A */ union UINT_FLOAT SQ_PGM_CF_OFFSET_VS ; /* 0xA234 */ GLboolean dirty; + int num_consts; + union UINT_FLOAT consts[R700_MAX_DX9_CONSTS][4]; } VS_STATE_STRUCT; typedef struct _GS_STATE_STRUCT diff --git a/src/mesa/drivers/dri/r600/r700_fragprog.c b/src/mesa/drivers/dri/r600/r700_fragprog.c index 6249bde6f18..098b420dfcc 100644 --- a/src/mesa/drivers/dri/r600/r700_fragprog.c +++ b/src/mesa/drivers/dri/r600/r700_fragprog.c @@ -270,7 +270,6 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) { context_t *context = R700_CONTEXT(ctx); R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw); - BATCH_LOCALS(&context->radeon); struct r700_fragment_program *fp = (struct r700_fragment_program *) (ctx->FragmentProgram._Current); r700_AssemblerBase *pAsm = &(fp->r700AsmCode); @@ -280,6 +279,7 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) unsigned int ui, i; unsigned int unNumOfReg; unsigned int unBit; + GLuint exportCount; if(GL_FALSE == fp->loaded) { @@ -305,8 +305,15 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) (context->chipobj.MemUse)(context, fp->shadercode.buf->id); */ + R600_STATECHANGE(context, ps); + + r700->ps.SQ_PGM_RESOURCES_PS.u32All = 0; + SETbit(r700->ps.SQ_PGM_RESOURCES_PS.u32All, PGM_RESOURCES__PRIME_CACHE_ON_DRAW_bit); + r700->ps.SQ_PGM_START_PS.u32All = 0; /* set from buffer obj */ + R600_STATECHANGE(context, spi); + unNumOfReg = fp->r700Shader.nRegs + 1; ui = (r700->SPI_PS_IN_CONTROL_0.u32All & NUM_INTERP_mask) / (1 << NUM_INTERP_shift); @@ -323,8 +330,8 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) ui = (unNumOfReg < ui) ? ui : unNumOfReg; - SETfield(r700->ps.SQ_PGM_RESOURCES_PS.u32All, ui, NUM_GPRS_shift, NUM_GPRS_mask); - + SETfield(r700->ps.SQ_PGM_RESOURCES_PS.u32All, ui, NUM_GPRS_shift, NUM_GPRS_mask); + CLEARbit(r700->ps.SQ_PGM_RESOURCES_PS.u32All, UNCACHED_FIRST_INST_bit); if(fp->r700Shader.uStackSize) /* we don't use branch for now, it should be zero. */ @@ -336,6 +343,8 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) SETfield(r700->ps.SQ_PGM_EXPORTS_PS.u32All, fp->r700Shader.exportMode, EXPORT_MODE_shift, EXPORT_MODE_mask); + R600_STATECHANGE(context, db); + if(fp->r700Shader.killIsUsed) { SETbit(r700->DB_SHADER_CONTROL.u32All, KILL_ENABLE_bit); @@ -347,42 +356,13 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) if(fp->r700Shader.depthIsExported) { - SETbit(r700->DB_SHADER_CONTROL.u32All, Z_EXPORT_ENABLE_bit); + SETbit(r700->DB_SHADER_CONTROL.u32All, Z_EXPORT_ENABLE_bit); } else { CLEARbit(r700->DB_SHADER_CONTROL.u32All, Z_EXPORT_ENABLE_bit); } - /* sent out shader constants. */ - paramList = fp->mesa_program.Base.Parameters; - - if(NULL != paramList) - { - _mesa_load_state_parameters(ctx, paramList); - - unNumParamData = paramList->NumParameters * 4; - - BEGIN_BATCH_NO_AUTOSTATE(2 + unNumParamData); - - R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_ALU_CONST, unNumParamData)); - - /* assembler map const from very beginning. */ - R600_OUT_BATCH(SQ_ALU_CONSTANT_PS_OFFSET * 4); - - unNumParamData = paramList->NumParameters; - - for(ui=0; uiParameterValues[ui][0]))); - R600_OUT_BATCH(*((unsigned int*)&(paramList->ParameterValues[ui][1]))); - R600_OUT_BATCH(*((unsigned int*)&(paramList->ParameterValues[ui][2]))); - R600_OUT_BATCH(*((unsigned int*)&(paramList->ParameterValues[ui][3]))); - } - END_BATCH(); - COMMIT_BATCH(); - } - // emit ps input map unBit = 1 << FRAG_ATTRIB_WPOS; if(mesa_fp->Base.InputsRead & unBit) @@ -449,6 +429,34 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) } } + R600_STATECHANGE(context, cb); + exportCount = (r700->ps.SQ_PGM_EXPORTS_PS.u32All & EXPORT_MODE_mask) / (1 << EXPORT_MODE_shift); + r700->CB_SHADER_CONTROL.u32All = (1 << exportCount) - 1; + + /* sent out shader constants. */ + paramList = fp->mesa_program.Base.Parameters; + + if(NULL != paramList) { + _mesa_load_state_parameters(ctx, paramList); + + if (paramList->NumParameters > R700_MAX_DX9_CONSTS) + return GL_FALSE; + + R600_STATECHANGE(context, ps_consts); + + r700->ps.num_consts = paramList->NumParameters; + + unNumParamData = paramList->NumParameters; + + for(ui=0; uips.consts[ui][0].f32All = paramList->ParameterValues[ui][0]; + r700->ps.consts[ui][1].f32All = paramList->ParameterValues[ui][1]; + r700->ps.consts[ui][2].f32All = paramList->ParameterValues[ui][2]; + r700->ps.consts[ui][3].f32All = paramList->ParameterValues[ui][3]; + } + } else + r700->ps.num_consts = 0; + return GL_TRUE; } diff --git a/src/mesa/drivers/dri/r600/r700_render.c b/src/mesa/drivers/dri/r600/r700_render.c index f0cd357c765..c8b405f5557 100644 --- a/src/mesa/drivers/dri/r600/r700_render.c +++ b/src/mesa/drivers/dri/r600/r700_render.c @@ -55,7 +55,6 @@ void r700WaitForIdle(context_t *context); void r700WaitForIdleClean(context_t *context); -void r700Start3D(context_t *context); GLboolean r700SendTextureState(context_t *context); static unsigned int r700PrimitiveType(int prim); void r600UpdateTextureState(GLcontext * ctx); @@ -116,39 +115,6 @@ void r700Start3D(context_t *context) r700WaitForIdleClean(context); } -static GLboolean r700SetupShaders(GLcontext * ctx) -{ - context_t *context = R700_CONTEXT(ctx); - - R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw); - - GLuint exportCount; - - r700->ps.SQ_PGM_RESOURCES_PS.u32All = 0; - r700->vs.SQ_PGM_RESOURCES_VS.u32All = 0; - - SETbit(r700->ps.SQ_PGM_RESOURCES_PS.u32All, PGM_RESOURCES__PRIME_CACHE_ON_DRAW_bit); - SETbit(r700->vs.SQ_PGM_RESOURCES_VS.u32All, PGM_RESOURCES__PRIME_CACHE_ON_DRAW_bit); - - r700SetupVertexProgram(ctx); - - r700SetupFragmentProgram(ctx); - - exportCount = (r700->ps.SQ_PGM_EXPORTS_PS.u32All & EXPORT_MODE_mask) / (1 << EXPORT_MODE_shift); - r700->CB_SHADER_CONTROL.u32All = (1 << exportCount) - 1; - - r600UpdateTextureState(ctx); - - r700SendFSState(context); // FIXME just a place holder for now - r700SendPSState(context); - r700SendVSState(context); - - r700SendTextureState(context); - r700SetupStreams(ctx); - - return GL_TRUE; -} - GLboolean r700SyncSurf(context_t *context, struct radeon_bo *pbo, uint32_t read_domain, @@ -333,7 +299,7 @@ static void r700RunRenderPrimitive(GLcontext * ctx, int start, int end, int prim } static GLboolean r700RunRender(GLcontext * ctx, - struct tnl_pipeline_stage *stage) + struct tnl_pipeline_stage *stage) { context_t *context = R700_CONTEXT(ctx); radeonContextPtr radeon = &context->radeon; @@ -347,12 +313,15 @@ static GLboolean r700RunRender(GLcontext * ctx, /* just an estimate, need to properly calculate this */ rcommonEnsureCmdBufSpace(&context->radeon, - radeon->hw.max_state_size + ind_count + 1000, __FUNCTION__); + radeon->hw.max_state_size + ind_count, __FUNCTION__); - r700Start3D(context); r700UpdateShaders(ctx); r700SetScissor(context); - r700SetupShaders(ctx); + r700SetupVertexProgram(ctx); + r700SetupFragmentProgram(ctx); + r600UpdateTextureState(ctx); + r700SetupStreams(ctx); + radeonEmitState(radeon); /* richard test code */ diff --git a/src/mesa/drivers/dri/r600/r700_state.c b/src/mesa/drivers/dri/r600/r700_state.c index 6b44cc0ceb6..75b2b72eb93 100644 --- a/src/mesa/drivers/dri/r600/r700_state.c +++ b/src/mesa/drivers/dri/r600/r700_state.c @@ -168,7 +168,6 @@ void r700UpdateViewportOffset(GLcontext * ctx) //------------------ void r700UpdateDrawBuffer(GLcontext * ctx) /* TODO */ //--------------------- { context_t *context = R700_CONTEXT(ctx); - R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw); R600_STATECHANGE(context, cb_target); R600_STATECHANGE(context, db_target); @@ -400,7 +399,7 @@ static void r700BlendColor(GLcontext * ctx, const GLfloat cf[4]) //------------- context_t *context = R700_CONTEXT(ctx); R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw); - R600_STATECHANGE(context, cb); + R600_STATECHANGE(context, blnd_clr); r700->CB_BLEND_RED.f32All = cf[0]; r700->CB_BLEND_GREEN.f32All = cf[1]; @@ -470,7 +469,7 @@ static void r700SetBlendState(GLcontext * ctx) int id = 0; uint32_t blend_reg = 0, eqn, eqnA; - R600_STATECHANGE(context, cb); + R600_STATECHANGE(context, blnd); if (RGBA_LOGICOP_ENABLED(ctx) || !ctx->Color.BlendEnabled) { SETfield(blend_reg, @@ -661,7 +660,7 @@ static void r700SetLogicOpState(GLcontext *ctx) context_t *context = R700_CONTEXT(ctx); R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&R700_CONTEXT(ctx)->hw); - R600_STATECHANGE(context, cb); + R600_STATECHANGE(context, blnd); if (RGBA_LOGICOP_ENABLED(ctx)) SETfield(r700->CB_COLOR_CONTROL.u32All, @@ -1024,7 +1023,7 @@ static void r700StencilFuncSeparate(GLcontext * ctx, GLenum face, //fixme //r300CatchStencilFallback(ctx); - R600_STATECHANGE(context, db); + R600_STATECHANGE(context, stencil); //front SETfield(r700->DB_STENCILREFMASK.u32All, ctx->Stencil.Ref[0], @@ -1055,7 +1054,7 @@ static void r700StencilMaskSeparate(GLcontext * ctx, GLenum face, GLuint mask) / //fixme //r300CatchStencilFallback(ctx); - R600_STATECHANGE(context, db); + R600_STATECHANGE(context, stencil); // front SETfield(r700->DB_STENCILREFMASK.u32All, ctx->Stencil.WriteMask[0], @@ -1215,7 +1214,7 @@ static void r700PolygonOffset(GLcontext * ctx, GLfloat factor, GLfloat units) // factor *= 12.0; - R600_STATECHANGE(context, su); + R600_STATECHANGE(context, poly); r700->PA_SU_POLY_OFFSET_FRONT_SCALE.f32All = factor; r700->PA_SU_POLY_OFFSET_FRONT_OFFSET.f32All = constant; @@ -1356,7 +1355,7 @@ void r700SetScissor(context_t *context) //--------------- y2 = rrb->dPriv->y + rrb->dPriv->h; } - R600_STATECHANGE(context, sc); + R600_STATECHANGE(context, scissor); /* window */ SETbit(r700->PA_SC_WINDOW_SCISSOR_TL.u32All, WINDOW_OFFSET_DISABLE_bit); @@ -1422,7 +1421,6 @@ static void r700SetRenderTarget(context_t *context, int id) rrb = radeon_get_colorbuffer(&context->radeon); if (!rrb || !rrb->bo) { - fprintf(stderr, "no rrb\n"); return; } @@ -1579,6 +1577,7 @@ static void r700InitSQConfig(GLcontext * ctx) case CHIP_FAMILY_RV610: case CHIP_FAMILY_RV620: case CHIP_FAMILY_RS780: + case CHIP_FAMILY_RS880: default: num_ps_gprs = 84; num_vs_gprs = 36; @@ -1661,6 +1660,7 @@ static void r700InitSQConfig(GLcontext * ctx) if ((context->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV610) || (context->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV620) || (context->radeon.radeonScreen->chip_family == CHIP_FAMILY_RS780) || + (context->radeon.radeonScreen->chip_family == CHIP_FAMILY_RS880) || (context->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV710)) CLEARbit(r700->sq_config.SQ_CONFIG.u32All, VC_ENABLE_bit); else diff --git a/src/mesa/drivers/dri/r600/r700_vertprog.c b/src/mesa/drivers/dri/r600/r700_vertprog.c index 31e71cdfa30..550594e9df8 100644 --- a/src/mesa/drivers/dri/r600/r700_vertprog.c +++ b/src/mesa/drivers/dri/r600/r700_vertprog.c @@ -336,7 +336,6 @@ GLboolean r700SetupVertexProgram(GLcontext * ctx) { context_t *context = R700_CONTEXT(ctx); R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw); - BATCH_LOCALS(&context->radeon); struct r700_vertex_program *vp = (struct r700_vertex_program *)ctx->VertexProgram._Current; @@ -368,8 +367,14 @@ GLboolean r700SetupVertexProgram(GLcontext * ctx) (context->chipobj.MemUse)(context, vp->shadercode.buf->id); */ - r700->vs.SQ_PGM_START_VS.u32All = 0; /* set from buffer object. */ - + R600_STATECHANGE(context, vs); + R600_STATECHANGE(context, fs); /* hack */ + + r700->vs.SQ_PGM_RESOURCES_VS.u32All = 0; + SETbit(r700->vs.SQ_PGM_RESOURCES_VS.u32All, PGM_RESOURCES__PRIME_CACHE_ON_DRAW_bit); + + r700->vs.SQ_PGM_START_VS.u32All = 0; /* set from buffer object. */ + SETfield(r700->vs.SQ_PGM_RESOURCES_VS.u32All, vp->r700Shader.nRegs + 1, NUM_GPRS_shift, NUM_GPRS_mask); @@ -379,9 +384,12 @@ GLboolean r700SetupVertexProgram(GLcontext * ctx) STACK_SIZE_shift, STACK_SIZE_mask); } - SETfield(r700->SPI_VS_OUT_CONFIG.u32All, vp->r700Shader.nParamExports ? (vp->r700Shader.nParamExports - 1) : 0, + R600_STATECHANGE(context, spi); + + SETfield(r700->SPI_VS_OUT_CONFIG.u32All, + vp->r700Shader.nParamExports ? (vp->r700Shader.nParamExports - 1) : 0, VS_EXPORT_COUNT_shift, VS_EXPORT_COUNT_mask); - SETfield(r700->SPI_PS_IN_CONTROL_0.u32All, vp->r700Shader.nParamExports, + SETfield(r700->SPI_PS_IN_CONTROL_0.u32All, vp->r700Shader.nParamExports, NUM_INTERP_shift, NUM_INTERP_mask); /* @@ -392,34 +400,26 @@ GLboolean r700SetupVertexProgram(GLcontext * ctx) /* sent out shader constants. */ paramList = vp->mesa_program.Base.Parameters; - if(NULL != paramList) - { - _mesa_load_state_parameters(ctx, paramList); + if(NULL != paramList) { + _mesa_load_state_parameters(ctx, paramList); - unNumParamData = paramList->NumParameters * 4; + if (paramList->NumParameters > R700_MAX_DX9_CONSTS) + return GL_FALSE; - BEGIN_BATCH_NO_AUTOSTATE(unNumParamData + 2); + R600_STATECHANGE(context, vs_consts); - R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_ALU_CONST, unNumParamData)); - /* assembler map const from very beginning. */ - R600_OUT_BATCH(SQ_ALU_CONSTANT_VS_OFFSET * 4); + r700->vs.num_consts = paramList->NumParameters; - unNumParamData = paramList->NumParameters; + unNumParamData = paramList->NumParameters; - for(ui=0; uiParameterValues[ui][0]))); - R600_OUT_BATCH(*((unsigned int*)&(paramList->ParameterValues[ui][1]))); - R600_OUT_BATCH(*((unsigned int*)&(paramList->ParameterValues[ui][2]))); - R600_OUT_BATCH(*((unsigned int*)&(paramList->ParameterValues[ui][3]))); - } - END_BATCH(); - COMMIT_BATCH(); - } + for(ui=0; uivs.consts[ui][0].f32All = paramList->ParameterValues[ui][0]; + r700->vs.consts[ui][1].f32All = paramList->ParameterValues[ui][1]; + r700->vs.consts[ui][2].f32All = paramList->ParameterValues[ui][2]; + r700->vs.consts[ui][3].f32All = paramList->ParameterValues[ui][3]; + } + } else + r700->vs.num_consts = 0; return GL_TRUE; } - - - - diff --git a/src/mesa/drivers/dri/radeon/radeon_bo_drm.h b/src/mesa/drivers/dri/radeon/radeon_bo_drm.h index 8789e3ab09b..71413716333 100644 --- a/src/mesa/drivers/dri/radeon/radeon_bo_drm.h +++ b/src/mesa/drivers/dri/radeon/radeon_bo_drm.h @@ -73,6 +73,7 @@ struct radeon_bo_funcs { uint32_t pitch); int (*bo_get_tiling)(struct radeon_bo *bo, uint32_t *tiling_flags, uint32_t *pitch); + int (*bo_is_busy)(struct radeon_bo *bo, uint32_t *domain); }; struct radeon_bo_manager { @@ -170,6 +171,15 @@ static inline int _radeon_bo_wait(struct radeon_bo *bo, return bo->bom->funcs->bo_wait(bo); } +static inline int _radeon_bo_is_busy(struct radeon_bo *bo, + uint32_t *domain, + const char *file, + const char *func, + int line) +{ + return bo->bom->funcs->bo_is_busy(bo, domain); +} + static inline int radeon_bo_set_tiling(struct radeon_bo *bo, uint32_t tiling_flags, uint32_t pitch) { @@ -203,5 +213,7 @@ static inline int radeon_bo_is_static(struct radeon_bo *bo) _radeon_bo_debug(bo, opcode, __FILE__, __FUNCTION__, __LINE__) #define radeon_bo_wait(bo) \ _radeon_bo_wait(bo, __FILE__, __func__, __LINE__) +#define radeon_bo_is_busy(bo, domain) \ + _radeon_bo_is_busy(bo, domain, __FILE__, __func__, __LINE__) #endif diff --git a/src/mesa/drivers/dri/radeon/radeon_bo_legacy.c b/src/mesa/drivers/dri/radeon/radeon_bo_legacy.c index b1cc155f71a..a10c6b73abc 100644 --- a/src/mesa/drivers/dri/radeon/radeon_bo_legacy.c +++ b/src/mesa/drivers/dri/radeon/radeon_bo_legacy.c @@ -542,6 +542,18 @@ static int bo_unmap(struct radeon_bo *bo) return 0; } +static int bo_is_busy(struct radeon_bo *bo, uint32_t *domain) +{ + *domain = 0; + if (bo->domains & RADEON_GEM_DOMAIN_GTT) + *domain = RADEON_GEM_DOMAIN_GTT; + else + *domain = RADEON_GEM_DOMAIN_CPU; + if (legacy_is_pending(bo)) + return -EBUSY; + else + return 0; +} static int bo_is_static(struct radeon_bo *bo) { @@ -559,6 +571,7 @@ static struct radeon_bo_funcs bo_legacy_funcs = { bo_is_static, NULL, NULL, + bo_is_busy }; static int bo_vram_validate(struct radeon_bo *bo, diff --git a/src/mesa/drivers/dri/radeon/radeon_chipset.h b/src/mesa/drivers/dri/radeon/radeon_chipset.h index a275c8fb143..0da1c0f9ee3 100644 --- a/src/mesa/drivers/dri/radeon/radeon_chipset.h +++ b/src/mesa/drivers/dri/radeon/radeon_chipset.h @@ -335,6 +335,12 @@ #define PCI_CHIP_RS780_9615 0x9615 #define PCI_CHIP_RS780_9616 0x9616 +#define PCI_CHIP_RS880_9710 0x9710 +#define PCI_CHIP_RS880_9711 0x9711 +#define PCI_CHIP_RS880_9712 0x9712 +#define PCI_CHIP_RS880_9713 0x9713 +#define PCI_CHIP_RS880_9714 0x9714 + #define PCI_CHIP_RV770_9440 0x9440 #define PCI_CHIP_RV770_9441 0x9441 #define PCI_CHIP_RV770_9442 0x9442 @@ -421,6 +427,7 @@ enum { CHIP_FAMILY_RV620, CHIP_FAMILY_RV635, CHIP_FAMILY_RS780, + CHIP_FAMILY_RS880, CHIP_FAMILY_RV770, CHIP_FAMILY_RV730, CHIP_FAMILY_RV710, diff --git a/src/mesa/drivers/dri/radeon/radeon_common_context.c b/src/mesa/drivers/dri/radeon/radeon_common_context.c index ad4584a2bde..ef296e491eb 100644 --- a/src/mesa/drivers/dri/radeon/radeon_common_context.c +++ b/src/mesa/drivers/dri/radeon/radeon_common_context.c @@ -92,6 +92,7 @@ static const char* get_chip_family_name(int chip_family) case CHIP_FAMILY_RV620: return "RV620"; case CHIP_FAMILY_RV635: return "RV635"; case CHIP_FAMILY_RS780: return "RS780"; + case CHIP_FAMILY_RS880: return "RS880"; case CHIP_FAMILY_RV770: return "RV770"; case CHIP_FAMILY_RV730: return "RV730"; case CHIP_FAMILY_RV710: return "RV710"; diff --git a/src/mesa/drivers/dri/radeon/radeon_cs_drm.h b/src/mesa/drivers/dri/radeon/radeon_cs_drm.h index ee403d173cc..ab4eca31a3c 100644 --- a/src/mesa/drivers/dri/radeon/radeon_cs_drm.h +++ b/src/mesa/drivers/dri/radeon/radeon_cs_drm.h @@ -201,6 +201,15 @@ static inline void radeon_cs_write_qword(struct radeon_cs *cs, uint64_t qword) } } +static inline void radeon_cs_write_table(struct radeon_cs *cs, void *data, uint32_t size) +{ + memcpy(cs->packets + cs->cdw, data, size * 4); + cs->cdw += size; + if (cs->section) { + cs->section_cdw += size; + } +} + static inline void radeon_cs_space_set_flush(struct radeon_cs *cs, void (*fn)(void *), void *data) { cs->space_flush_fn = fn; diff --git a/src/mesa/drivers/dri/radeon/radeon_dma.c b/src/mesa/drivers/dri/radeon/radeon_dma.c index 7e8f50e3df2..f306befec47 100644 --- a/src/mesa/drivers/dri/radeon/radeon_dma.c +++ b/src/mesa/drivers/dri/radeon/radeon_dma.c @@ -30,6 +30,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ +#include #include "radeon_common.h" #include "main/simple_list.h" @@ -302,7 +303,13 @@ void radeonReturnDmaRegion(radeonContextPtr rmesa, int return_bytes) static int radeon_bo_is_idle(struct radeon_bo* bo) { - return bo->cref == 1; + uint32_t domain; + int ret = radeon_bo_is_busy(bo, &domain); + if (ret == -EINVAL) { + WARN_ONCE("Your libdrm or kernel doesn't have support for busy query.\n" + "This may cause small performance drop for you.\n"); + } + return ret != -EBUSY; } void radeonReleaseDmaRegions(radeonContextPtr rmesa) diff --git a/src/mesa/drivers/dri/radeon/radeon_queryobj.c b/src/mesa/drivers/dri/radeon/radeon_queryobj.c index 70251946dff..7eef4faaf61 100644 --- a/src/mesa/drivers/dri/radeon/radeon_queryobj.c +++ b/src/mesa/drivers/dri/radeon/radeon_queryobj.c @@ -32,7 +32,19 @@ #define DDEBUG 0 -#define PAGE_SIZE 4096 +static int radeonQueryIsFlushed(GLcontext *ctx, struct gl_query_object *q) +{ + radeonContextPtr radeon = RADEON_CONTEXT(ctx); + struct radeon_query_object *tmp, *query = (struct radeon_query_object *)q; + + foreach(tmp, &radeon->query.not_flushed_head) { + if (tmp == query) { + return 0; + } + } + + return 1; +} static void radeonQueryGetResult(GLcontext *ctx, struct gl_query_object *q) { @@ -86,22 +98,11 @@ static void radeonDeleteQuery(GLcontext *ctx, struct gl_query_object *q) static void radeonWaitQuery(GLcontext *ctx, struct gl_query_object *q) { - radeonContextPtr radeon = RADEON_CONTEXT(ctx); - struct radeon_query_object *tmp, *query = (struct radeon_query_object *)q; + struct radeon_query_object *query = (struct radeon_query_object *)q; /* If the cmdbuf with packets for this query hasn't been flushed yet, do it now */ - { - GLboolean found = GL_FALSE; - foreach(tmp, &radeon->query.not_flushed_head) { - if (tmp == query) { - found = GL_TRUE; - break; - } - } - - if (found) - ctx->Driver.Flush(ctx); - } + if (!radeonQueryIsFlushed(ctx, q)) + ctx->Driver.Flush(ctx); if (DDEBUG) fprintf(stderr, "%s: query id %d, bo %p, offset %d\n", __FUNCTION__, q->Id, query->bo, query->curr_offset); @@ -124,7 +125,7 @@ static void radeonBeginQuery(GLcontext *ctx, struct gl_query_object *q) radeon->dma.flush(radeon->glCtx); if (!query->bo) { - query->bo = radeon_bo_open(radeon->radeonScreen->bom, 0, PAGE_SIZE, PAGE_SIZE, RADEON_GEM_DOMAIN_GTT, 0); + query->bo = radeon_bo_open(radeon->radeonScreen->bom, 0, RADEON_QUERY_PAGE_SIZE, RADEON_QUERY_PAGE_SIZE, RADEON_GEM_DOMAIN_GTT, 0); } query->curr_offset = 0; @@ -168,16 +169,32 @@ static void radeonEndQuery(GLcontext *ctx, struct gl_query_object *q) radeon->query.current = NULL; } -/** - * TODO: - * should check if bo is idle, bo there's no interface to do it - * just wait for result now - */ static void radeonCheckQuery(GLcontext *ctx, struct gl_query_object *q) { if (DDEBUG) fprintf(stderr, "%s: query id %d\n", __FUNCTION__, q->Id); +#ifdef DRM_RADEON_GEM_BUSY + radeonContextPtr radeon = RADEON_CONTEXT(ctx); + + if (radeon->radeonScreen->kernel_mm) { + struct radeon_query_object *query = (struct radeon_query_object *)q; + uint32_t domain; + + /* Need to perform a flush, as per ARB_occlusion_query spec */ + if (!radeonQueryIsFlushed(ctx, q)) { + ctx->Driver.Flush(ctx); + } + + if (radeon_bo_is_busy(query->bo, &domain) == 0) { + radeonQueryGetResult(ctx, q); + query->Base.Ready = GL_TRUE; + } + } else { + radeonWaitQuery(ctx, q); + } +#else radeonWaitQuery(ctx, q); +#endif } void radeonInitQueryObjFunctions(struct dd_function_table *functions) diff --git a/src/mesa/drivers/dri/radeon/radeon_screen.c b/src/mesa/drivers/dri/radeon/radeon_screen.c index e28543d855a..10afe527d3d 100644 --- a/src/mesa/drivers/dri/radeon/radeon_screen.c +++ b/src/mesa/drivers/dri/radeon/radeon_screen.c @@ -853,6 +853,14 @@ static int radeon_set_screen_flags(radeonScreenPtr screen, int device_id) screen->chip_family = CHIP_FAMILY_RS780; screen->chip_flags = RADEON_CHIPSET_TCL; break; + case PCI_CHIP_RS880_9710: + case PCI_CHIP_RS880_9711: + case PCI_CHIP_RS880_9712: + case PCI_CHIP_RS880_9713: + case PCI_CHIP_RS880_9714: + screen->chip_family = CHIP_FAMILY_RS880; + screen->chip_flags = RADEON_CHIPSET_TCL; + break; case PCI_CHIP_RV770_9440: case PCI_CHIP_RV770_9441: diff --git a/src/mesa/drivers/dri/radeon/radeon_state_init.c b/src/mesa/drivers/dri/radeon/radeon_state_init.c index e70dbbe06f2..bed3935130b 100644 --- a/src/mesa/drivers/dri/radeon/radeon_state_init.c +++ b/src/mesa/drivers/dri/radeon/radeon_state_init.c @@ -876,7 +876,7 @@ void radeonInitState( r100ContextPtr rmesa ) rmesa->hw.glt.emit = vec_emit; rmesa->hw.eye.emit = vec_emit; - for (i = 0; i <= 6; i++) + for (i = 0; i < 6; i++) rmesa->hw.mat[i].emit = vec_emit; for (i = 0; i < 8; i++) diff --git a/src/mesa/shader/.gitignore b/src/mesa/shader/.gitignore new file mode 100644 index 00000000000..086fd9a705c --- /dev/null +++ b/src/mesa/shader/.gitignore @@ -0,0 +1 @@ +program_parse.output diff --git a/src/mesa/shader/Makefile b/src/mesa/shader/Makefile new file mode 100644 index 00000000000..400a543bdab --- /dev/null +++ b/src/mesa/shader/Makefile @@ -0,0 +1,7 @@ +all: program_parse.tab.c lex.yy.c + +program_parse.tab.c program_parse.tab.h: program_parse.y + bison -v -d $< + +lex.yy.c: program_lexer.l + flex --never-interactive $< diff --git a/src/mesa/shader/arbprogparse.c b/src/mesa/shader/arbprogparse.c index 381445c0e84..05ee4f563eb 100644 --- a/src/mesa/shader/arbprogparse.c +++ b/src/mesa/shader/arbprogparse.c @@ -63,3868 +63,7 @@ having three separate program parameter arrays. #include "prog_parameter.h" #include "prog_statevars.h" #include "prog_instruction.h" - -/** - * This is basically a union of the vertex_program and fragment_program - * structs that we can use to parse the program into - * - * XXX we can probably get rid of this entirely someday. - */ -struct arb_program -{ - struct gl_program Base; - - GLuint Position; /* Just used for error reporting while parsing */ - GLuint MajorVersion; - GLuint MinorVersion; - - /* ARB_vertex_progmra options */ - GLboolean HintPositionInvariant; - - /* ARB_fragment_progmra options */ - GLenum PrecisionOption; /* GL_DONT_CARE, GL_NICEST or GL_FASTEST */ - GLenum FogOption; /* GL_NONE, GL_LINEAR, GL_EXP or GL_EXP2 */ - - /* ARB_fragment_program specifics */ - GLbitfield TexturesUsed[MAX_TEXTURE_IMAGE_UNITS]; - GLbitfield ShadowSamplers; - GLuint NumAluInstructions; - GLuint NumTexInstructions; - GLuint NumTexIndirections; - - GLboolean UsesKill; -}; - - - -/* TODO: - * Fragment Program Stuff: - * ----------------------------------------------------- - * - * - things from Michal's email - * + overflow on atoi - * + not-overflowing floats (don't use parse_integer..) - * + can remove range checking in arbparse.c - * - * - check all limits of number of various variables - * + parameters - * - * - test! test! test! - * - * Vertex Program Stuff: - * ----------------------------------------------------- - * - Optimize param array usage and count limits correctly, see spec, - * section 2.14.3.7 - * + Record if an array is reference absolutly or relatively (or both) - * + For absolute arrays, store a bitmap of accesses - * + For single parameters, store an access flag - * + After parsing, make a parameter cleanup and merging pass, where - * relative arrays are layed out first, followed by abs arrays, and - * finally single state. - * + Remap offsets for param src and dst registers - * + Now we can properly count parameter usage - * - * - Multiple state binding errors in param arrays (see spec, just before - * section 2.14.3.3) - * - grep for XXX - * - * Mesa Stuff - * ----------------------------------------------------- - * - User clipping planes vs. PositionInvariant - * - Is it sufficient to just multiply by the mvp to transform in the - * PositionInvariant case? Or do we need something more involved? - * - * - vp_src swizzle is GLubyte, fp_src swizzle is GLuint - * - fetch state listed in program_parameters list - * + WTF should this go??? - * + currently in nvvertexec.c and s_nvfragprog.c - * - * - allow for multiple address registers (and fetch address regs properly) - * - * Cosmetic Stuff - * ----------------------------------------------------- - * - remove any leftover unused grammer.c stuff (dict_ ?) - * - fix grammer.c error handling so its not static - * - #ifdef around stuff pertaining to extentions - * - * Outstanding Questions: - * ----------------------------------------------------- - * - ARB_matrix_palette / ARB_vertex_blend -- not supported - * what gets hacked off because of this: - * + VERTEX_ATTRIB_MATRIXINDEX - * + VERTEX_ATTRIB_WEIGHT - * + MATRIX_MODELVIEW - * + MATRIX_PALETTE - * - * - When can we fetch env/local params from their own register files, and - * when to we have to fetch them into the main state register file? - * (think arrays) - * - * Grammar Changes: - * ----------------------------------------------------- - */ - -/* Changes since moving the file to shader directory - -2004-III-4 ------------------------------------------------------------ -- added #include "grammar_mesa.h" -- removed grammar specific code part (it resides now in grammar.c) -- added GL_ARB_fragment_program_shadow tokens -- modified #include "arbparse_syn.h" -- major changes inside _mesa_parse_arb_program() -- check the program string for '\0' characters -- copy the program string to a one-byte-longer location to have - it null-terminated -- position invariance test (not writing to result.position) moved - to syntax part -*/ - -typedef GLubyte *production; - - -/** - * This is the text describing the rules to parse the grammar - */ -LONGSTRING static char arb_grammar_text[] = -#include "arbprogram_syn.h" -; - -/** - * These should match up with the values defined in arbprogram.syn - */ - -/* - Changes: - - changed and merged V_* and F_* opcode values to OP_*. - - added GL_ARB_fragment_program_shadow specific tokens (michal) -*/ -#define REVISION 0x0a - -/* program type */ -#define FRAGMENT_PROGRAM 0x01 -#define VERTEX_PROGRAM 0x02 - -/* program section */ -#define OPTION 0x01 -#define INSTRUCTION 0x02 -#define DECLARATION 0x03 -#define END 0x04 - -/* GL_ARB_fragment_program option */ -#define ARB_PRECISION_HINT_FASTEST 0x00 -#define ARB_PRECISION_HINT_NICEST 0x01 -#define ARB_FOG_EXP 0x02 -#define ARB_FOG_EXP2 0x03 -#define ARB_FOG_LINEAR 0x04 - -/* GL_ARB_vertex_program option */ -#define ARB_POSITION_INVARIANT 0x05 - -/* GL_ARB_fragment_program_shadow option */ -#define ARB_FRAGMENT_PROGRAM_SHADOW 0x06 - -/* GL_ARB_draw_buffers option */ -#define ARB_DRAW_BUFFERS 0x07 - -/* GL_MESA_texture_array option */ -#define MESA_TEXTURE_ARRAY 0x08 - -/* GL_ARB_fragment_program instruction class */ -#define OP_ALU_INST 0x00 -#define OP_TEX_INST 0x01 - -/* GL_ARB_vertex_program instruction class */ -/* OP_ALU_INST */ - -/* GL_ARB_fragment_program instruction type */ -#define OP_ALU_VECTOR 0x00 -#define OP_ALU_SCALAR 0x01 -#define OP_ALU_BINSC 0x02 -#define OP_ALU_BIN 0x03 -#define OP_ALU_TRI 0x04 -#define OP_ALU_SWZ 0x05 -#define OP_TEX_SAMPLE 0x06 -#define OP_TEX_KIL 0x07 - -/* GL_ARB_vertex_program instruction type */ -#define OP_ALU_ARL 0x08 -/* OP_ALU_VECTOR */ -/* OP_ALU_SCALAR */ -/* OP_ALU_BINSC */ -/* OP_ALU_BIN */ -/* OP_ALU_TRI */ -/* OP_ALU_SWZ */ - -/* GL_ARB_fragment_program instruction code */ -#define OP_ABS 0x00 -#define OP_ABS_SAT 0x1B -#define OP_FLR 0x09 -#define OP_FLR_SAT 0x26 -#define OP_FRC 0x0A -#define OP_FRC_SAT 0x27 -#define OP_LIT 0x0C -#define OP_LIT_SAT 0x2A -#define OP_MOV 0x11 -#define OP_MOV_SAT 0x30 -#define OP_COS 0x1F -#define OP_COS_SAT 0x20 -#define OP_EX2 0x07 -#define OP_EX2_SAT 0x25 -#define OP_LG2 0x0B -#define OP_LG2_SAT 0x29 -#define OP_RCP 0x14 -#define OP_RCP_SAT 0x33 -#define OP_RSQ 0x15 -#define OP_RSQ_SAT 0x34 -#define OP_SIN 0x38 -#define OP_SIN_SAT 0x39 -#define OP_SCS 0x35 -#define OP_SCS_SAT 0x36 -#define OP_POW 0x13 -#define OP_POW_SAT 0x32 -#define OP_ADD 0x01 -#define OP_ADD_SAT 0x1C -#define OP_DP3 0x03 -#define OP_DP3_SAT 0x21 -#define OP_DP4 0x04 -#define OP_DP4_SAT 0x22 -#define OP_DPH 0x05 -#define OP_DPH_SAT 0x23 -#define OP_DST 0x06 -#define OP_DST_SAT 0x24 -#define OP_MAX 0x0F -#define OP_MAX_SAT 0x2E -#define OP_MIN 0x10 -#define OP_MIN_SAT 0x2F -#define OP_MUL 0x12 -#define OP_MUL_SAT 0x31 -#define OP_SGE 0x16 -#define OP_SGE_SAT 0x37 -#define OP_SLT 0x17 -#define OP_SLT_SAT 0x3A -#define OP_SUB 0x18 -#define OP_SUB_SAT 0x3B -#define OP_XPD 0x1A -#define OP_XPD_SAT 0x43 -#define OP_CMP 0x1D -#define OP_CMP_SAT 0x1E -#define OP_LRP 0x2B -#define OP_LRP_SAT 0x2C -#define OP_MAD 0x0E -#define OP_MAD_SAT 0x2D -#define OP_SWZ 0x19 -#define OP_SWZ_SAT 0x3C -#define OP_TEX 0x3D -#define OP_TEX_SAT 0x3E -#define OP_TXB 0x3F -#define OP_TXB_SAT 0x40 -#define OP_TXP 0x41 -#define OP_TXP_SAT 0x42 -#define OP_KIL 0x28 - -/* GL_ARB_vertex_program instruction code */ -#define OP_ARL 0x02 -/* OP_ABS */ -/* OP_FLR */ -/* OP_FRC */ -/* OP_LIT */ -/* OP_MOV */ -/* OP_EX2 */ -#define OP_EXP 0x08 -/* OP_LG2 */ -#define OP_LOG 0x0D -/* OP_RCP */ -/* OP_RSQ */ -/* OP_POW */ -/* OP_ADD */ -/* OP_DP3 */ -/* OP_DP4 */ -/* OP_DPH */ -/* OP_DST */ -/* OP_MAX */ -/* OP_MIN */ -/* OP_MUL */ -/* OP_SGE */ -/* OP_SLT */ -/* OP_SUB */ -/* OP_XPD */ -/* OP_MAD */ -/* OP_SWZ */ - -/* fragment attribute binding */ -#define FRAGMENT_ATTRIB_COLOR 0x01 -#define FRAGMENT_ATTRIB_TEXCOORD 0x02 -#define FRAGMENT_ATTRIB_FOGCOORD 0x03 -#define FRAGMENT_ATTRIB_POSITION 0x04 - -/* vertex attribute binding */ -#define VERTEX_ATTRIB_POSITION 0x01 -#define VERTEX_ATTRIB_WEIGHT 0x02 -#define VERTEX_ATTRIB_NORMAL 0x03 -#define VERTEX_ATTRIB_COLOR 0x04 -#define VERTEX_ATTRIB_FOGCOORD 0x05 -#define VERTEX_ATTRIB_TEXCOORD 0x06 -#define VERTEX_ATTRIB_MATRIXINDEX 0x07 -#define VERTEX_ATTRIB_GENERIC 0x08 - -/* fragment result binding */ -#define FRAGMENT_RESULT_COLOR 0x01 -#define FRAGMENT_RESULT_DEPTH 0x02 - -/* vertex result binding */ -#define VERTEX_RESULT_POSITION 0x01 -#define VERTEX_RESULT_COLOR 0x02 -#define VERTEX_RESULT_FOGCOORD 0x03 -#define VERTEX_RESULT_POINTSIZE 0x04 -#define VERTEX_RESULT_TEXCOORD 0x05 - -/* texture target */ -#define TEXTARGET_1D 0x01 -#define TEXTARGET_2D 0x02 -#define TEXTARGET_3D 0x03 -#define TEXTARGET_RECT 0x04 -#define TEXTARGET_CUBE 0x05 -/* GL_ARB_fragment_program_shadow */ -#define TEXTARGET_SHADOW1D 0x06 -#define TEXTARGET_SHADOW2D 0x07 -#define TEXTARGET_SHADOWRECT 0x08 -/* GL_MESA_texture_array */ -#define TEXTARGET_1D_ARRAY 0x09 -#define TEXTARGET_2D_ARRAY 0x0a -#define TEXTARGET_SHADOW1D_ARRAY 0x0b -#define TEXTARGET_SHADOW2D_ARRAY 0x0c - -/* face type */ -#define FACE_FRONT 0x00 -#define FACE_BACK 0x01 - -/* color type */ -#define COLOR_PRIMARY 0x00 -#define COLOR_SECONDARY 0x01 - -/* component */ -#define COMPONENT_X 0x00 -#define COMPONENT_Y 0x01 -#define COMPONENT_Z 0x02 -#define COMPONENT_W 0x03 -#define COMPONENT_0 0x04 -#define COMPONENT_1 0x05 - -/* array index type */ -#define ARRAY_INDEX_ABSOLUTE 0x00 -#define ARRAY_INDEX_RELATIVE 0x01 - -/* matrix name */ -#define MATRIX_MODELVIEW 0x01 -#define MATRIX_PROJECTION 0x02 -#define MATRIX_MVP 0x03 -#define MATRIX_TEXTURE 0x04 -#define MATRIX_PALETTE 0x05 -#define MATRIX_PROGRAM 0x06 - -/* matrix modifier */ -#define MATRIX_MODIFIER_IDENTITY 0x00 -#define MATRIX_MODIFIER_INVERSE 0x01 -#define MATRIX_MODIFIER_TRANSPOSE 0x02 -#define MATRIX_MODIFIER_INVTRANS 0x03 - -/* constant type */ -#define CONSTANT_SCALAR 0x01 -#define CONSTANT_VECTOR 0x02 - -/* program param type */ -#define PROGRAM_PARAM_ENV 0x01 -#define PROGRAM_PARAM_LOCAL 0x02 - -/* register type */ -#define REGISTER_ATTRIB 0x01 -#define REGISTER_PARAM 0x02 -#define REGISTER_RESULT 0x03 -#define REGISTER_ESTABLISHED_NAME 0x04 - -/* param binding */ -#define PARAM_NULL 0x00 -#define PARAM_ARRAY_ELEMENT 0x01 -#define PARAM_STATE_ELEMENT 0x02 -#define PARAM_PROGRAM_ELEMENT 0x03 -#define PARAM_PROGRAM_ELEMENTS 0x04 -#define PARAM_CONSTANT 0x05 - -/* param state property */ -#define STATE_MATERIAL_PARSER 0x01 -#define STATE_LIGHT_PARSER 0x02 -#define STATE_LIGHT_MODEL 0x03 -#define STATE_LIGHT_PROD 0x04 -#define STATE_FOG 0x05 -#define STATE_MATRIX_ROWS 0x06 -/* GL_ARB_fragment_program */ -#define STATE_TEX_ENV 0x07 -#define STATE_DEPTH 0x08 -/* GL_ARB_vertex_program */ -#define STATE_TEX_GEN 0x09 -#define STATE_CLIP_PLANE 0x0A -#define STATE_POINT 0x0B - -/* state material property */ -#define MATERIAL_AMBIENT 0x01 -#define MATERIAL_DIFFUSE 0x02 -#define MATERIAL_SPECULAR 0x03 -#define MATERIAL_EMISSION 0x04 -#define MATERIAL_SHININESS 0x05 - -/* state light property */ -#define LIGHT_AMBIENT 0x01 -#define LIGHT_DIFFUSE 0x02 -#define LIGHT_SPECULAR 0x03 -#define LIGHT_POSITION 0x04 -#define LIGHT_ATTENUATION 0x05 -#define LIGHT_HALF 0x06 -#define LIGHT_SPOT_DIRECTION 0x07 - -/* state light model property */ -#define LIGHT_MODEL_AMBIENT 0x01 -#define LIGHT_MODEL_SCENECOLOR 0x02 - -/* state light product property */ -#define LIGHT_PROD_AMBIENT 0x01 -#define LIGHT_PROD_DIFFUSE 0x02 -#define LIGHT_PROD_SPECULAR 0x03 - -/* state texture environment property */ -#define TEX_ENV_COLOR 0x01 - -/* state texture generation coord property */ -#define TEX_GEN_EYE 0x01 -#define TEX_GEN_OBJECT 0x02 - -/* state fog property */ -#define FOG_COLOR 0x01 -#define FOG_PARAMS 0x02 - -/* state depth property */ -#define DEPTH_RANGE 0x01 - -/* state point parameters property */ -#define POINT_SIZE 0x01 -#define POINT_ATTENUATION 0x02 - -/* declaration */ -#define ATTRIB 0x01 -#define PARAM 0x02 -#define TEMP 0x03 -#define OUTPUT 0x04 -#define ALIAS 0x05 -/* GL_ARB_vertex_program */ -#define ADDRESS 0x06 - -/*----------------------------------------------------------------------- - * From here on down is the semantic checking portion - * - */ - -/** - * Variable Table Handling functions - */ -typedef enum -{ - vt_none, - vt_address, - vt_attrib, - vt_param, - vt_temp, - vt_output, - vt_alias -} var_type; - - -/** - * Setting an explicit field for each of the binding properties is a bit - * wasteful of space, but it should be much more clear when reading later on.. - */ -struct var_cache -{ - const GLubyte *name; /* don't free() - no need */ - var_type type; - GLuint address_binding; /* The index of the address register we should - * be using */ - GLuint attrib_binding; /* For type vt_attrib, see nvfragprog.h for values */ - GLuint attrib_is_generic; /* If the attrib was specified through a generic - * vertex attrib */ - GLuint temp_binding; /* The index of the temp register we are to use */ - GLuint output_binding; /* Output/result register number */ - struct var_cache *alias_binding; /* For type vt_alias, points to the var_cache entry - * that this is aliased to */ - GLuint param_binding_type; /* {PROGRAM_STATE_VAR, PROGRAM_LOCAL_PARAM, - * PROGRAM_ENV_PARAM} */ - GLuint param_binding_begin; /* This is the offset into the program_parameter_list where - * the tokens representing our bound state (or constants) - * start */ - GLuint param_binding_length; /* This is how many entries in the the program_parameter_list - * we take up with our state tokens or constants. Note that - * this is _not_ the same as the number of param registers - * we eventually use */ - GLuint swizzle; /**< swizzle to access this variable */ - struct var_cache *next; -}; - -static GLvoid -var_cache_create (struct var_cache **va) -{ - *va = (struct var_cache *) _mesa_malloc (sizeof (struct var_cache)); - if (*va) { - (**va).name = NULL; - (**va).type = vt_none; - (**va).attrib_binding = ~0; - (**va).attrib_is_generic = 0; - (**va).temp_binding = ~0; - (**va).output_binding = ~0; - (**va).param_binding_type = ~0; - (**va).param_binding_begin = ~0; - (**va).param_binding_length = ~0; - (**va).alias_binding = NULL; - (**va).swizzle = SWIZZLE_XYZW; - (**va).next = NULL; - } -} - -static GLvoid -var_cache_destroy (struct var_cache **va) -{ - if (*va) { - var_cache_destroy (&(**va).next); - _mesa_free (*va); - *va = NULL; - } -} - -static GLvoid -var_cache_append (struct var_cache **va, struct var_cache *nv) -{ - if (*va) - var_cache_append (&(**va).next, nv); - else - *va = nv; -} - -static struct var_cache * -var_cache_find (struct var_cache *va, const GLubyte * name) -{ - /*struct var_cache *first = va;*/ - - while (va) { - if (!_mesa_strcmp ( (const char*) name, (const char*) va->name)) { - if (va->type == vt_alias) - return va->alias_binding; - return va; - } - - va = va->next; - } - - return NULL; -} - - - -/** - * Called when an error is detected while parsing/compiling a program. - * Sets the ctx->Program.ErrorString field to descript and records a - * GL_INVALID_OPERATION error. - * \param position position of error in program string - * \param descrip verbose error description - */ -static void -program_error(GLcontext *ctx, GLint position, const char *descrip) -{ - if (descrip) { - const char *prefix = "glProgramString(", *suffix = ")"; - char *str = (char *) _mesa_malloc(_mesa_strlen(descrip) + - _mesa_strlen(prefix) + - _mesa_strlen(suffix) + 1); - if (str) { - _mesa_sprintf(str, "%s%s%s", prefix, descrip, suffix); - _mesa_error(ctx, GL_INVALID_OPERATION, str); - _mesa_free(str); - } - } - _mesa_set_program_error(ctx, position, descrip); -} - - -/** - * As above, but with an extra string parameter for more info. - */ -static void -program_error2(GLcontext *ctx, GLint position, const char *descrip, - const char *var) -{ - if (descrip) { - const char *prefix = "glProgramString(", *suffix = ")"; - char *str = (char *) _mesa_malloc(_mesa_strlen(descrip) + - _mesa_strlen(": ") + - _mesa_strlen(var) + - _mesa_strlen(prefix) + - _mesa_strlen(suffix) + 1); - if (str) { - _mesa_sprintf(str, "%s%s: %s%s", prefix, descrip, var, suffix); - _mesa_error(ctx, GL_INVALID_OPERATION, str); - _mesa_free(str); - } - } - { - char *str = (char *) _mesa_malloc(_mesa_strlen(descrip) + - _mesa_strlen(": ") + - _mesa_strlen(var) + 1); - if (str) { - _mesa_sprintf(str, "%s: %s", descrip, var); - } - _mesa_set_program_error(ctx, position, str); - if (str) { - _mesa_free(str); - } - } -} - - - -/** - * constructs an integer from 4 GLubytes in LE format - */ -static GLuint -parse_position (const GLubyte ** inst) -{ - GLuint value; - - value = (GLuint) (*(*inst)++); - value += (GLuint) (*(*inst)++) * 0x100; - value += (GLuint) (*(*inst)++) * 0x10000; - value += (GLuint) (*(*inst)++) * 0x1000000; - - return value; -} - -/** - * This will, given a string, lookup the string as a variable name in the - * var cache. If the name is found, the var cache node corresponding to the - * var name is returned. If it is not found, a new entry is allocated - * - * \param I Points into the binary array where the string identifier begins - * \param found 1 if the string was found in the var_cache, 0 if it was allocated - * \return The location on the var_cache corresponding the the string starting at I - */ -static struct var_cache * -parse_string (const GLubyte ** inst, struct var_cache **vc_head, - struct arb_program *Program, GLuint * found) -{ - const GLubyte *i = *inst; - struct var_cache *va = NULL; - (void) Program; - - *inst += _mesa_strlen ((char *) i) + 1; - - va = var_cache_find (*vc_head, i); - - if (va) { - *found = 1; - return va; - } - - *found = 0; - var_cache_create (&va); - va->name = (const GLubyte *) i; - - var_cache_append (vc_head, va); - - return va; -} - -static char * -parse_string_without_adding (const GLubyte ** inst, struct arb_program *Program) -{ - const GLubyte *i = *inst; - (void) Program; - - *inst += _mesa_strlen ((char *) i) + 1; - - return (char *) i; -} - -/** - * \return -1 if we parse '-', return 1 otherwise - */ -static GLint -parse_sign (const GLubyte ** inst) -{ - /*return *(*inst)++ != '+'; */ - - if (**inst == '-') { - (*inst)++; - return -1; - } - else if (**inst == '+') { - (*inst)++; - return 1; - } - - return 1; -} - -/** - * parses and returns signed integer - */ -static GLint -parse_integer (const GLubyte ** inst, struct arb_program *Program) -{ - GLint sign; - GLint value; - - /* check if *inst points to '+' or '-' - * if yes, grab the sign and increment *inst - */ - sign = parse_sign (inst); - - /* now check if *inst points to 0 - * if yes, increment the *inst and return the default value - */ - if (**inst == 0) { - (*inst)++; - return 0; - } - - /* parse the integer as you normally would do it */ - value = _mesa_atoi (parse_string_without_adding (inst, Program)); - - /* now, after terminating 0 there is a position - * to parse it - parse_position() - */ - Program->Position = parse_position (inst); - - return value * sign; -} - -/** - Accumulate this string of digits, and return them as - a large integer represented in floating point (for range). - If scale is not NULL, also accumulates a power-of-ten - integer scale factor that represents the number of digits - in the string. -*/ -static GLdouble -parse_float_string(const GLubyte ** inst, struct arb_program *Program, GLdouble *scale) -{ - GLdouble value = 0.0; - GLdouble oscale = 1.0; - - if (**inst == 0) { /* this string of digits is empty-- do nothing */ - (*inst)++; - } - else { /* nonempty string-- parse out the digits */ - while (**inst >= '0' && **inst <= '9') { - GLubyte digit = *((*inst)++); - value = value * 10.0 + (GLint) (digit - '0'); - oscale *= 10.0; - } - assert(**inst == 0); /* integer string should end with 0 */ - (*inst)++; /* skip over terminating 0 */ - Program->Position = parse_position(inst); /* skip position (from integer) */ - } - if (scale) - *scale = oscale; - return value; -} - -/** - Parse an unsigned floating-point number from this stream of tokenized - characters. Example floating-point formats supported: - 12.34 - 12 - 0.34 - .34 - 12.34e-4 - */ -static GLfloat -parse_float (const GLubyte ** inst, struct arb_program *Program) -{ - GLint exponent; - GLdouble whole, fraction, fracScale = 1.0; - - whole = parse_float_string(inst, Program, 0); - fraction = parse_float_string(inst, Program, &fracScale); - - /* Parse signed exponent */ - exponent = parse_integer(inst, Program); /* This is the exponent */ - - /* Assemble parts of floating-point number: */ - return (GLfloat) ((whole + fraction / fracScale) * - _mesa_pow(10.0, (GLfloat) exponent)); -} - - -/** - */ -static GLfloat -parse_signed_float (const GLubyte ** inst, struct arb_program *Program) -{ - GLint sign = parse_sign (inst); - GLfloat value = parse_float (inst, Program); - return value * sign; -} - -/** - * This picks out a constant value from the parsed array. The constant vector is r - * returned in the *values array, which should be of length 4. - * - * \param values - return the vector constant values. - * \param size - returns the number elements in valuesOut [1..4] - */ -static GLvoid -parse_constant(const GLubyte ** inst, GLfloat *values, GLint *size, - struct arb_program *Program, - GLboolean use) -{ - GLuint components, i; - - switch (*(*inst)++) { - case CONSTANT_SCALAR: - if (use == GL_TRUE) { - values[0] = - values[1] = - values[2] = values[3] = parse_float (inst, Program); - } - else { - values[0] = - values[1] = - values[2] = values[3] = parse_signed_float (inst, Program); - } - *size = 1; - break; - case CONSTANT_VECTOR: - values[0] = values[1] = values[2] = 0; - values[3] = 1; - components = *(*inst)++; - for (i = 0; i < components; i++) { - values[i] = parse_signed_float (inst, Program); - } - *size = 4; - break; - default: - _mesa_problem(NULL, "unexpected case in parse_constant()"); - values[0] = 0.0F; - *size = 0; - } -} - -/** - * \param offset The offset from the address register that we should - * address - * - * \return 0 on sucess, 1 on error - */ -static GLuint -parse_relative_offset(GLcontext *ctx, const GLubyte **inst, - struct arb_program *Program, GLint *offset) -{ - (void) ctx; - *offset = parse_integer(inst, Program); - return 0; -} - -/** - * \param color 0 if color type is primary, 1 if color type is secondary - * \return 0 on sucess, 1 on error - */ -static GLuint -parse_color_type (GLcontext * ctx, const GLubyte ** inst, struct arb_program *Program, - GLint * color) -{ - (void) ctx; (void) Program; - *color = *(*inst)++ != COLOR_PRIMARY; - return 0; -} - -/** - * Get an integer corresponding to a generic vertex attribute. - * - * \return 0 on sucess, 1 on error - */ -static GLuint -parse_generic_attrib_num(GLcontext *ctx, const GLubyte ** inst, - struct arb_program *Program, GLuint *attrib) -{ - GLint i = parse_integer(inst, Program); - - if ((i < 0) || (i >= MAX_VERTEX_GENERIC_ATTRIBS)) - { - program_error(ctx, Program->Position, - "Invalid generic vertex attribute index"); - return 1; - } - - *attrib = (GLuint) i; - - return 0; -} - - -/** - * \param color The index of the color buffer to write into - * \return 0 on sucess, 1 on error - */ -static GLuint -parse_output_color_num (GLcontext * ctx, const GLubyte ** inst, - struct arb_program *Program, GLuint * color) -{ - GLint i = parse_integer (inst, Program); - - if ((i < 0) || (i >= (int)ctx->Const.MaxDrawBuffers)) { - *color = 0; - program_error(ctx, Program->Position, "Invalid draw buffer index"); - return 1; - } - - *color = (GLuint) i; - return 0; -} - - -/** - * Validate the index of a texture coordinate - * - * \param coord The texture unit index - * \return 0 on sucess, 1 on error - */ -static GLuint -parse_texcoord_num (GLcontext * ctx, const GLubyte ** inst, - struct arb_program *Program, GLuint * coord) -{ - GLint i = parse_integer (inst, Program); - - if ((i < 0) || (i >= (int)ctx->Const.MaxTextureCoordUnits)) { - program_error(ctx, Program->Position, "Invalid texture coordinate index"); - return 1; - } - - *coord = (GLuint) i; - return 0; -} - - -/** - * Validate the index of a texture image unit - * - * \param coord The texture unit index - * \return 0 on sucess, 1 on error - */ -static GLuint -parse_teximage_num (GLcontext * ctx, const GLubyte ** inst, - struct arb_program *Program, GLuint * coord) -{ - GLint i = parse_integer (inst, Program); - - if ((i < 0) || (i >= (int)ctx->Const.MaxTextureImageUnits)) { - char s[100]; - _mesa_snprintf(s, sizeof(s), "Invalid texture image index %d (%u is max)", - i, ctx->Const.MaxTextureImageUnits); - program_error(ctx, Program->Position, s); - return 1; - } - - *coord = (GLuint) i; - return 0; -} - - -/** - * \param coord The weight index - * \return 0 on sucess, 1 on error - */ -static GLuint -parse_weight_num (GLcontext * ctx, const GLubyte ** inst, struct arb_program *Program, - GLint * coord) -{ - *coord = parse_integer (inst, Program); - - if ((*coord < 0) || (*coord >= 1)) { - program_error(ctx, Program->Position, "Invalid weight index"); - return 1; - } - - return 0; -} - -/** - * \param coord The clip plane index - * \return 0 on sucess, 1 on error - */ -static GLuint -parse_clipplane_num (GLcontext * ctx, const GLubyte ** inst, - struct arb_program *Program, GLint * coord) -{ - *coord = parse_integer (inst, Program); - - if ((*coord < 0) || (*coord >= (GLint) ctx->Const.MaxClipPlanes)) { - program_error(ctx, Program->Position, "Invalid clip plane index"); - return 1; - } - - return 0; -} - - -/** - * \return 0 on front face, 1 on back face - */ -static GLuint -parse_face_type (const GLubyte ** inst) -{ - switch (*(*inst)++) { - case FACE_FRONT: - return 0; - - case FACE_BACK: - return 1; - } - return 0; -} - - -/** - * Given a matrix and a modifier token on the binary array, return tokens - * that _mesa_fetch_state() [program.c] can understand. - * - * \param matrix - the matrix we are talking about - * \param matrix_idx - the index of the matrix we have (for texture & program matricies) - * \param matrix_modifier - the matrix modifier (trans, inv, etc) - * \return 0 on sucess, 1 on failure - */ -static GLuint -parse_matrix (GLcontext * ctx, const GLubyte ** inst, struct arb_program *Program, - GLint * matrix, GLint * matrix_idx, GLint * matrix_modifier) -{ - GLubyte mat = *(*inst)++; - - *matrix_idx = 0; - - switch (mat) { - case MATRIX_MODELVIEW: - *matrix = STATE_MODELVIEW_MATRIX; - *matrix_idx = parse_integer (inst, Program); - if (*matrix_idx > 0) { - program_error(ctx, Program->Position, - "ARB_vertex_blend not supported"); - return 1; - } - break; - - case MATRIX_PROJECTION: - *matrix = STATE_PROJECTION_MATRIX; - break; - - case MATRIX_MVP: - *matrix = STATE_MVP_MATRIX; - break; - - case MATRIX_TEXTURE: - *matrix = STATE_TEXTURE_MATRIX; - *matrix_idx = parse_integer (inst, Program); - if (*matrix_idx >= (GLint) ctx->Const.MaxTextureUnits) { - program_error(ctx, Program->Position, "Invalid Texture Unit"); - /* bad *matrix_id */ - return 1; - } - break; - - /* This is not currently supported (ARB_matrix_palette) */ - case MATRIX_PALETTE: - *matrix_idx = parse_integer (inst, Program); - program_error(ctx, Program->Position, - "ARB_matrix_palette not supported"); - return 1; - break; - - case MATRIX_PROGRAM: - *matrix = STATE_PROGRAM_MATRIX; - *matrix_idx = parse_integer (inst, Program); - if (*matrix_idx >= (GLint) ctx->Const.MaxProgramMatrices) { - program_error(ctx, Program->Position, "Invalid Program Matrix"); - /* bad *matrix_idx */ - return 1; - } - break; - } - - switch (*(*inst)++) { - case MATRIX_MODIFIER_IDENTITY: - *matrix_modifier = 0; - break; - case MATRIX_MODIFIER_INVERSE: - *matrix_modifier = STATE_MATRIX_INVERSE; - break; - case MATRIX_MODIFIER_TRANSPOSE: - *matrix_modifier = STATE_MATRIX_TRANSPOSE; - break; - case MATRIX_MODIFIER_INVTRANS: - *matrix_modifier = STATE_MATRIX_INVTRANS; - break; - } - - return 0; -} - - -/** - * This parses a state string (rather, the binary version of it) into - * a 6-token sequence as described in _mesa_fetch_state() [program.c] - * - * \param inst - the start in the binary arry to start working from - * \param state_tokens - the storage for the 6-token state description - * \return - 0 on sucess, 1 on error - */ -static GLuint -parse_state_single_item (GLcontext * ctx, const GLubyte ** inst, - struct arb_program *Program, - gl_state_index state_tokens[STATE_LENGTH]) -{ - GLubyte token = *(*inst)++; - - switch (token) { - case STATE_MATERIAL_PARSER: - state_tokens[0] = STATE_MATERIAL; - state_tokens[1] = parse_face_type (inst); - switch (*(*inst)++) { - case MATERIAL_AMBIENT: - state_tokens[2] = STATE_AMBIENT; - break; - case MATERIAL_DIFFUSE: - state_tokens[2] = STATE_DIFFUSE; - break; - case MATERIAL_SPECULAR: - state_tokens[2] = STATE_SPECULAR; - break; - case MATERIAL_EMISSION: - state_tokens[2] = STATE_EMISSION; - break; - case MATERIAL_SHININESS: - state_tokens[2] = STATE_SHININESS; - break; - } - break; - - case STATE_LIGHT_PARSER: - state_tokens[0] = STATE_LIGHT; - state_tokens[1] = parse_integer (inst, Program); - - /* Check the value of state_tokens[1] against the # of lights */ - if (state_tokens[1] >= (GLint) ctx->Const.MaxLights) { - program_error(ctx, Program->Position, "Invalid Light Number"); - /* bad state_tokens[1] */ - return 1; - } - - switch (*(*inst)++) { - case LIGHT_AMBIENT: - state_tokens[2] = STATE_AMBIENT; - break; - case LIGHT_DIFFUSE: - state_tokens[2] = STATE_DIFFUSE; - break; - case LIGHT_SPECULAR: - state_tokens[2] = STATE_SPECULAR; - break; - case LIGHT_POSITION: - state_tokens[2] = STATE_POSITION; - break; - case LIGHT_ATTENUATION: - state_tokens[2] = STATE_ATTENUATION; - break; - case LIGHT_HALF: - state_tokens[2] = STATE_HALF_VECTOR; - break; - case LIGHT_SPOT_DIRECTION: - state_tokens[2] = STATE_SPOT_DIRECTION; - break; - } - break; - - case STATE_LIGHT_MODEL: - switch (*(*inst)++) { - case LIGHT_MODEL_AMBIENT: - state_tokens[0] = STATE_LIGHTMODEL_AMBIENT; - break; - case LIGHT_MODEL_SCENECOLOR: - state_tokens[0] = STATE_LIGHTMODEL_SCENECOLOR; - state_tokens[1] = parse_face_type (inst); - break; - } - break; - - case STATE_LIGHT_PROD: - state_tokens[0] = STATE_LIGHTPROD; - state_tokens[1] = parse_integer (inst, Program); - - /* Check the value of state_tokens[1] against the # of lights */ - if (state_tokens[1] >= (GLint) ctx->Const.MaxLights) { - program_error(ctx, Program->Position, "Invalid Light Number"); - /* bad state_tokens[1] */ - return 1; - } - - state_tokens[2] = parse_face_type (inst); - switch (*(*inst)++) { - case LIGHT_PROD_AMBIENT: - state_tokens[3] = STATE_AMBIENT; - break; - case LIGHT_PROD_DIFFUSE: - state_tokens[3] = STATE_DIFFUSE; - break; - case LIGHT_PROD_SPECULAR: - state_tokens[3] = STATE_SPECULAR; - break; - } - break; - - - case STATE_FOG: - switch (*(*inst)++) { - case FOG_COLOR: - state_tokens[0] = STATE_FOG_COLOR; - break; - case FOG_PARAMS: - state_tokens[0] = STATE_FOG_PARAMS; - break; - } - break; - - case STATE_TEX_ENV: - state_tokens[1] = parse_integer (inst, Program); - switch (*(*inst)++) { - case TEX_ENV_COLOR: - state_tokens[0] = STATE_TEXENV_COLOR; - break; - } - break; - - case STATE_TEX_GEN: - { - GLuint type, coord; - - state_tokens[0] = STATE_TEXGEN; - /*state_tokens[1] = parse_integer (inst, Program);*/ /* Texture Unit */ - - if (parse_texcoord_num (ctx, inst, Program, &coord)) - return 1; - state_tokens[1] = coord; - - /* EYE or OBJECT */ - type = *(*inst)++; - - /* 0 - s, 1 - t, 2 - r, 3 - q */ - coord = *(*inst)++; - - if (type == TEX_GEN_EYE) { - switch (coord) { - case COMPONENT_X: - state_tokens[2] = STATE_TEXGEN_EYE_S; - break; - case COMPONENT_Y: - state_tokens[2] = STATE_TEXGEN_EYE_T; - break; - case COMPONENT_Z: - state_tokens[2] = STATE_TEXGEN_EYE_R; - break; - case COMPONENT_W: - state_tokens[2] = STATE_TEXGEN_EYE_Q; - break; - default: - _mesa_problem(ctx, "bad texgen component in " - "parse_state_single_item()"); - } - } - else { - switch (coord) { - case COMPONENT_X: - state_tokens[2] = STATE_TEXGEN_OBJECT_S; - break; - case COMPONENT_Y: - state_tokens[2] = STATE_TEXGEN_OBJECT_T; - break; - case COMPONENT_Z: - state_tokens[2] = STATE_TEXGEN_OBJECT_R; - break; - case COMPONENT_W: - state_tokens[2] = STATE_TEXGEN_OBJECT_Q; - break; - default: - _mesa_problem(ctx, "bad texgen component in " - "parse_state_single_item()"); - } - } - } - break; - - case STATE_DEPTH: - switch (*(*inst)++) { - case DEPTH_RANGE: - state_tokens[0] = STATE_DEPTH_RANGE; - break; - } - break; - - case STATE_CLIP_PLANE: - state_tokens[0] = STATE_CLIPPLANE; - if (parse_clipplane_num (ctx, inst, Program, - (GLint *) &state_tokens[1])) - return 1; - break; - - case STATE_POINT: - switch (*(*inst)++) { - case POINT_SIZE: - state_tokens[0] = STATE_POINT_SIZE; - break; - - case POINT_ATTENUATION: - state_tokens[0] = STATE_POINT_ATTENUATION; - break; - } - break; - - /* XXX: I think this is the correct format for a matrix row */ - case STATE_MATRIX_ROWS: - if (parse_matrix(ctx, inst, Program, - (GLint *) &state_tokens[0], - (GLint *) &state_tokens[1], - (GLint *) &state_tokens[4])) - return 1; - - state_tokens[2] = parse_integer (inst, Program); /* The first row to grab */ - - if ((**inst) != 0) { /* Either the last row, 0 */ - state_tokens[3] = parse_integer (inst, Program); - if (state_tokens[3] < state_tokens[2]) { - program_error(ctx, Program->Position, - "Second matrix index less than the first"); - /* state_tokens[4] vs. state_tokens[3] */ - return 1; - } - } - else { - state_tokens[3] = state_tokens[2]; - (*inst)++; - } - break; - } - - return 0; -} - -/** - * This parses a state string (rather, the binary version of it) into - * a 6-token similar for the state fetching code in program.c - * - * One might ask, why fetch these parameters into just like you fetch - * state when they are already stored in other places? - * - * Because of array offsets -> We can stick env/local parameters in the - * middle of a parameter array and then index someplace into the array - * when we execute. - * - * One optimization might be to only do this for the cases where the - * env/local parameters end up inside of an array, and leave the - * single parameters (or arrays of pure env/local pareameters) in their - * respective register files. - * - * For ENV parameters, the format is: - * state_tokens[0] = STATE_FRAGMENT_PROGRAM / STATE_VERTEX_PROGRAM - * state_tokens[1] = STATE_ENV - * state_tokens[2] = the parameter index - * - * for LOCAL parameters, the format is: - * state_tokens[0] = STATE_FRAGMENT_PROGRAM / STATE_VERTEX_PROGRAM - * state_tokens[1] = STATE_LOCAL - * state_tokens[2] = the parameter index - * - * \param inst - the start in the binary arry to start working from - * \param state_tokens - the storage for the 6-token state description - * \return - 0 on sucess, 1 on failure - */ -static GLuint -parse_program_single_item (GLcontext * ctx, const GLubyte ** inst, - struct arb_program *Program, - gl_state_index state_tokens[STATE_LENGTH]) -{ - if (Program->Base.Target == GL_FRAGMENT_PROGRAM_ARB) - state_tokens[0] = STATE_FRAGMENT_PROGRAM; - else - state_tokens[0] = STATE_VERTEX_PROGRAM; - - - switch (*(*inst)++) { - case PROGRAM_PARAM_ENV: - state_tokens[1] = STATE_ENV; - state_tokens[2] = parse_integer (inst, Program); - - /* Check state_tokens[2] against the number of ENV parameters available */ - if (((Program->Base.Target == GL_FRAGMENT_PROGRAM_ARB) && - (state_tokens[2] >= (GLint) ctx->Const.FragmentProgram.MaxEnvParams)) - || - ((Program->Base.Target == GL_VERTEX_PROGRAM_ARB) && - (state_tokens[2] >= (GLint) ctx->Const.VertexProgram.MaxEnvParams))) { - program_error(ctx, Program->Position, - "Invalid Program Env Parameter"); - /* bad state_tokens[2] */ - return 1; - } - - break; - - case PROGRAM_PARAM_LOCAL: - state_tokens[1] = STATE_LOCAL; - state_tokens[2] = parse_integer (inst, Program); - - /* Check state_tokens[2] against the number of LOCAL parameters available */ - if (((Program->Base.Target == GL_FRAGMENT_PROGRAM_ARB) && - (state_tokens[2] >= (GLint) ctx->Const.FragmentProgram.MaxLocalParams)) - || - ((Program->Base.Target == GL_VERTEX_PROGRAM_ARB) && - (state_tokens[2] >= (GLint) ctx->Const.VertexProgram.MaxLocalParams))) { - program_error(ctx, Program->Position, - "Invalid Program Local Parameter"); - /* bad state_tokens[2] */ - return 1; - } - break; - } - - return 0; -} - -/** - * For ARB_vertex_program, programs are not allowed to use both an explicit - * vertex attribute and a generic vertex attribute corresponding to the same - * state. See section 2.14.3.1 of the GL_ARB_vertex_program spec. - * - * This will walk our var_cache and make sure that nobody does anything fishy. - * - * \return 0 on sucess, 1 on error - */ -static GLuint -generic_attrib_check(struct var_cache *vc_head) -{ - int a; - struct var_cache *curr; - GLboolean explicitAttrib[MAX_VERTEX_GENERIC_ATTRIBS], - genericAttrib[MAX_VERTEX_GENERIC_ATTRIBS]; - - for (a=0; atype == vt_attrib) { - if (curr->attrib_is_generic) { - GLuint attr = (curr->attrib_binding == 0) - ? 0 : (curr->attrib_binding - VERT_ATTRIB_GENERIC0); - assert(attr < MAX_VERTEX_GENERIC_ATTRIBS); - genericAttrib[attr] = GL_TRUE; - } - else { - assert(curr->attrib_binding < MAX_VERTEX_GENERIC_ATTRIBS); - explicitAttrib[ curr->attrib_binding ] = GL_TRUE; - } - } - - curr = curr->next; - } - - for (a=0; aBase.Target == GL_FRAGMENT_PROGRAM_ARB) { - switch (*(*inst)++) { - case FRAGMENT_ATTRIB_COLOR: - { - GLint coord; - err = parse_color_type (ctx, inst, Program, &coord); - *inputReg = FRAG_ATTRIB_COL0 + coord; - } - break; - case FRAGMENT_ATTRIB_TEXCOORD: - { - GLuint texcoord = 0; - err = parse_texcoord_num (ctx, inst, Program, &texcoord); - *inputReg = FRAG_ATTRIB_TEX0 + texcoord; - } - break; - case FRAGMENT_ATTRIB_FOGCOORD: - *inputReg = FRAG_ATTRIB_FOGC; - break; - case FRAGMENT_ATTRIB_POSITION: - *inputReg = FRAG_ATTRIB_WPOS; - break; - default: - err = 1; - break; - } - } - else { - switch (*(*inst)++) { - case VERTEX_ATTRIB_POSITION: - *inputReg = VERT_ATTRIB_POS; - break; - - case VERTEX_ATTRIB_WEIGHT: - { - GLint weight; - err = parse_weight_num (ctx, inst, Program, &weight); - *inputReg = VERT_ATTRIB_WEIGHT; -#if 1 - /* hack for Warcraft (see bug 8060) */ - _mesa_warning(ctx, "Application error: vertex program uses 'vertex.weight' but GL_ARB_vertex_blend not supported."); - break; -#else - program_error(ctx, Program->Position, - "ARB_vertex_blend not supported"); - return 1; -#endif - } - - case VERTEX_ATTRIB_NORMAL: - *inputReg = VERT_ATTRIB_NORMAL; - break; - - case VERTEX_ATTRIB_COLOR: - { - GLint color; - err = parse_color_type (ctx, inst, Program, &color); - if (color) { - *inputReg = VERT_ATTRIB_COLOR1; - } - else { - *inputReg = VERT_ATTRIB_COLOR0; - } - } - break; - - case VERTEX_ATTRIB_FOGCOORD: - *inputReg = VERT_ATTRIB_FOG; - break; - - case VERTEX_ATTRIB_TEXCOORD: - { - GLuint unit = 0; - err = parse_texcoord_num (ctx, inst, Program, &unit); - *inputReg = VERT_ATTRIB_TEX0 + unit; - } - break; - - case VERTEX_ATTRIB_MATRIXINDEX: - /* Not supported at this time */ - { - const char *msg = "ARB_palette_matrix not supported"; - parse_integer (inst, Program); - program_error(ctx, Program->Position, msg); - } - return 1; - - case VERTEX_ATTRIB_GENERIC: - { - GLuint attrib; - err = parse_generic_attrib_num(ctx, inst, Program, &attrib); - if (!err) { - *is_generic = 1; - /* Add VERT_ATTRIB_GENERIC0 here because ARB_vertex_program's - * attributes do not alias the conventional vertex - * attributes. - */ - if (attrib > 0) - *inputReg = attrib + VERT_ATTRIB_GENERIC0; - else - *inputReg = 0; - } - } - break; - - default: - err = 1; - break; - } - } - - if (err) { - program_error(ctx, Program->Position, "Bad attribute binding"); - } - - return err; -} - - -/** - * This translates between a binary token for an output variable type - * and the mesa token for the same thing. - * - * \param inst The parsed tokens - * \param outputReg Returned index/number of the output register, - * one of the VERT_RESULT_* or FRAG_RESULT_* values. - */ -static GLuint -parse_result_binding(GLcontext *ctx, const GLubyte **inst, - GLuint *outputReg, struct arb_program *Program) -{ - const GLubyte token = *(*inst)++; - - switch (token) { - case FRAGMENT_RESULT_COLOR: - if (Program->Base.Target == GL_FRAGMENT_PROGRAM_ARB) { - GLuint out_color; - - /* This gets result of the color buffer we're supposed to - * draw into. This pertains to GL_ARB_draw_buffers. - */ - parse_output_color_num(ctx, inst, Program, &out_color); - ASSERT(out_color < MAX_DRAW_BUFFERS); - *outputReg = FRAG_RESULT_COLOR; - } - else { - /* for vtx programs, this is VERTEX_RESULT_POSITION */ - *outputReg = VERT_RESULT_HPOS; - } - break; - - case FRAGMENT_RESULT_DEPTH: - if (Program->Base.Target == GL_FRAGMENT_PROGRAM_ARB) { - /* for frag programs, this is FRAGMENT_RESULT_DEPTH */ - *outputReg = FRAG_RESULT_DEPTH; - } - else { - /* for vtx programs, this is VERTEX_RESULT_COLOR */ - GLint color_type; - GLuint face_type = parse_face_type(inst); - GLint err = parse_color_type(ctx, inst, Program, &color_type); - if (err) - return 1; - - if (face_type) { - /* back face */ - if (color_type) { - *outputReg = VERT_RESULT_BFC1; /* secondary color */ - } - else { - *outputReg = VERT_RESULT_BFC0; /* primary color */ - } - } - else { - /* front face */ - if (color_type) { - *outputReg = VERT_RESULT_COL1; /* secondary color */ - } - /* primary color */ - else { - *outputReg = VERT_RESULT_COL0; /* primary color */ - } - } - } - break; - - case VERTEX_RESULT_FOGCOORD: - *outputReg = VERT_RESULT_FOGC; - break; - - case VERTEX_RESULT_POINTSIZE: - *outputReg = VERT_RESULT_PSIZ; - break; - - case VERTEX_RESULT_TEXCOORD: - { - GLuint unit; - if (parse_texcoord_num (ctx, inst, Program, &unit)) - return 1; - *outputReg = VERT_RESULT_TEX0 + unit; - } - break; - } - - Program->Base.OutputsWritten |= (1 << *outputReg); - - return 0; -} - - -/** - * This handles the declaration of ATTRIB variables - * - * XXX: Still needs - * parse_vert_attrib_binding(), or something like that - * - * \return 0 on sucess, 1 on error - */ -static GLint -parse_attrib (GLcontext * ctx, const GLubyte ** inst, struct var_cache **vc_head, - struct arb_program *Program) -{ - GLuint found; - struct var_cache *attrib_var; - - attrib_var = parse_string (inst, vc_head, Program, &found); - Program->Position = parse_position (inst); - if (found) { - program_error2(ctx, Program->Position, - "Duplicate variable declaration", - (char *) attrib_var->name); - return 1; - } - - attrib_var->type = vt_attrib; - - if (parse_attrib_binding(ctx, inst, Program, &attrib_var->attrib_binding, - &attrib_var->attrib_is_generic)) - return 1; - - if (generic_attrib_check(*vc_head)) { - program_error(ctx, Program->Position, - "Cannot use both a generic vertex attribute " - "and a specific attribute of the same type"); - return 1; - } - - Program->Base.NumAttributes++; - return 0; -} - -/** - * \param use -- TRUE if we're called when declaring implicit parameters, - * FALSE if we're declaraing variables. This has to do with - * if we get a signed or unsigned float for scalar constants - */ -static GLuint -parse_param_elements (GLcontext * ctx, const GLubyte ** inst, - struct var_cache *param_var, - struct arb_program *Program, GLboolean use) -{ - GLint idx; - GLuint err = 0; - gl_state_index state_tokens[STATE_LENGTH] = {0, 0, 0, 0, 0}; - - GLubyte token = *(*inst)++; - - switch (token) { - case PARAM_STATE_ELEMENT: - if (parse_state_single_item (ctx, inst, Program, state_tokens)) - return 1; - - /* If we adding STATE_MATRIX that has multiple rows, we need to - * unroll it and call _mesa_add_state_reference() for each row - */ - if ((state_tokens[0] == STATE_MODELVIEW_MATRIX || - state_tokens[0] == STATE_PROJECTION_MATRIX || - state_tokens[0] == STATE_MVP_MATRIX || - state_tokens[0] == STATE_TEXTURE_MATRIX || - state_tokens[0] == STATE_PROGRAM_MATRIX) - && (state_tokens[2] != state_tokens[3])) { - GLint row; - const GLint first_row = state_tokens[2]; - const GLint last_row = state_tokens[3]; - - for (row = first_row; row <= last_row; row++) { - state_tokens[2] = state_tokens[3] = row; - - idx = _mesa_add_state_reference(Program->Base.Parameters, - state_tokens); - if (param_var->param_binding_begin == ~0U) - param_var->param_binding_begin = idx; - param_var->param_binding_length++; - } - } - else { - idx = _mesa_add_state_reference(Program->Base.Parameters, - state_tokens); - if (param_var->param_binding_begin == ~0U) - param_var->param_binding_begin = idx; - param_var->param_binding_length++; - } - break; - - case PARAM_PROGRAM_ELEMENT: - if (parse_program_single_item (ctx, inst, Program, state_tokens)) - return 1; - idx = _mesa_add_state_reference (Program->Base.Parameters, state_tokens); - if (param_var->param_binding_begin == ~0U) - param_var->param_binding_begin = idx; - param_var->param_binding_length++; - - /* Check if there is more: 0 -> we're done, else its an integer */ - if (**inst) { - GLuint out_of_range, new_idx; - GLuint start_idx = state_tokens[2] + 1; - GLuint end_idx = parse_integer (inst, Program); - - out_of_range = 0; - if (Program->Base.Target == GL_FRAGMENT_PROGRAM_ARB) { - if (((state_tokens[1] == STATE_ENV) - && (end_idx >= ctx->Const.FragmentProgram.MaxEnvParams)) - || ((state_tokens[1] == STATE_LOCAL) - && (end_idx >= - ctx->Const.FragmentProgram.MaxLocalParams))) - out_of_range = 1; - } - else { - if (((state_tokens[1] == STATE_ENV) - && (end_idx >= ctx->Const.VertexProgram.MaxEnvParams)) - || ((state_tokens[1] == STATE_LOCAL) - && (end_idx >= - ctx->Const.VertexProgram.MaxLocalParams))) - out_of_range = 1; - } - if (out_of_range) { - program_error(ctx, Program->Position, - "Invalid Program Parameter"); /*end_idx*/ - return 1; - } - - for (new_idx = start_idx; new_idx <= end_idx; new_idx++) { - state_tokens[2] = new_idx; - idx = _mesa_add_state_reference(Program->Base.Parameters, - state_tokens); - param_var->param_binding_length++; - } - } - else { - (*inst)++; - } - break; - - case PARAM_CONSTANT: - /* parsing something like {1.0, 2.0, 3.0, 4.0} */ - { - GLfloat const_values[4]; - GLint size; - parse_constant(inst, const_values, &size, Program, use); - if (param_var->name[0] == ' ') { - /* this is an unnamed constant */ - idx = _mesa_add_unnamed_constant(Program->Base.Parameters, - const_values, size, - ¶m_var->swizzle); - } - else { - /* named parameter/constant */ - idx = _mesa_add_named_constant(Program->Base.Parameters, - (char *) param_var->name, - const_values, size); - } - if (param_var->param_binding_begin == ~0U) - param_var->param_binding_begin = idx; - param_var->param_binding_type = PROGRAM_STATE_VAR; - /* Note: when we reference this parameter in an instruction later, - * we'll check if it's really a constant/immediate and set the - * instruction register type appropriately. - */ - param_var->param_binding_length++; - } - break; - - default: - program_error(ctx, Program->Position, - "Unexpected token (in parse_param_elements())"); - return 1; - } - - Program->Base.NumParameters = Program->Base.Parameters->NumParameters; - - /* Make sure we haven't blown past our parameter limits */ - if (((Program->Base.Target == GL_VERTEX_PROGRAM_ARB) && - (Program->Base.NumParameters > - ctx->Const.VertexProgram.MaxLocalParams)) - || ((Program->Base.Target == GL_FRAGMENT_PROGRAM_ARB) - && (Program->Base.NumParameters > - ctx->Const.FragmentProgram.MaxLocalParams))) { - program_error(ctx, Program->Position, "Too many parameter variables"); - return 1; - } - - return err; -} - - -/** - * This picks out PARAM program parameter bindings. - * - * XXX: This needs to be stressed & tested - * - * \return 0 on sucess, 1 on error - */ -static GLuint -parse_param (GLcontext * ctx, const GLubyte ** inst, struct var_cache **vc_head, - struct arb_program *Program) -{ - GLuint found, err; - GLint specified_length; - struct var_cache *param_var; - - err = 0; - param_var = parse_string (inst, vc_head, Program, &found); - Program->Position = parse_position (inst); - - if (found) { - program_error2(ctx, Program->Position, - "Duplicate variable declaration", - (char *) param_var->name); - return 1; - } - - specified_length = parse_integer (inst, Program); - - if (specified_length < 0) { - program_error(ctx, Program->Position, "Negative parameter array length"); - return 1; - } - - param_var->type = vt_param; - param_var->param_binding_length = 0; - - /* Right now, everything is shoved into the main state register file. - * - * In the future, it would be nice to leave things ENV/LOCAL params - * in their respective register files, if possible - */ - param_var->param_binding_type = PROGRAM_STATE_VAR; - - /* Remember to: - * * - add each guy to the parameter list - * * - increment the param_var->param_binding_len - * * - store the param_var->param_binding_begin for the first one - * * - compare the actual len to the specified len at the end - */ - while (**inst != PARAM_NULL) { - if (parse_param_elements (ctx, inst, param_var, Program, GL_FALSE)) - return 1; - } - - /* Test array length here! */ - if (specified_length) { - if (specified_length != (int)param_var->param_binding_length) { - program_error(ctx, Program->Position, - "Declared parameter array length does not match parameter list"); - return 1; - } - } - - (*inst)++; - - return 0; -} - -/** - * - */ -static GLuint -parse_param_use (GLcontext * ctx, const GLubyte ** inst, struct var_cache **vc_head, - struct arb_program *Program, struct var_cache **new_var) -{ - struct var_cache *param_var; - - /* First, insert a dummy entry into the var_cache */ - var_cache_create (¶m_var); - param_var->name = (const GLubyte *) " "; - param_var->type = vt_param; - - param_var->param_binding_length = 0; - /* Don't fill in binding_begin; We use the default value of -1 - * to tell if its already initialized, elsewhere. - * - * param_var->param_binding_begin = 0; - */ - param_var->param_binding_type = PROGRAM_STATE_VAR; - - var_cache_append (vc_head, param_var); - - /* Then fill it with juicy parameter goodness */ - if (parse_param_elements (ctx, inst, param_var, Program, GL_TRUE)) - return 1; - - *new_var = param_var; - - return 0; -} - - -/** - * This handles the declaration of TEMP variables - * - * \return 0 on sucess, 1 on error - */ -static GLuint -parse_temp (GLcontext * ctx, const GLubyte ** inst, struct var_cache **vc_head, - struct arb_program *Program) -{ - GLuint found; - struct var_cache *temp_var; - - while (**inst != 0) { - temp_var = parse_string (inst, vc_head, Program, &found); - Program->Position = parse_position (inst); - if (found) { - program_error2(ctx, Program->Position, - "Duplicate variable declaration", - (char *) temp_var->name); - return 1; - } - - temp_var->type = vt_temp; - - if (((Program->Base.Target == GL_FRAGMENT_PROGRAM_ARB) && - (Program->Base.NumTemporaries >= - ctx->Const.FragmentProgram.MaxTemps)) - || ((Program->Base.Target == GL_VERTEX_PROGRAM_ARB) - && (Program->Base.NumTemporaries >= - ctx->Const.VertexProgram.MaxTemps))) { - program_error(ctx, Program->Position, - "Too many TEMP variables declared"); - return 1; - } - - temp_var->temp_binding = Program->Base.NumTemporaries; - Program->Base.NumTemporaries++; - } - (*inst)++; - - return 0; -} - -/** - * This handles variables of the OUTPUT variety - * - * \return 0 on sucess, 1 on error - */ -static GLuint -parse_output (GLcontext * ctx, const GLubyte ** inst, struct var_cache **vc_head, - struct arb_program *Program) -{ - GLuint found; - struct var_cache *output_var; - GLuint err; - - output_var = parse_string (inst, vc_head, Program, &found); - Program->Position = parse_position (inst); - if (found) { - program_error2(ctx, Program->Position, - "Duplicate variable declaration", - (char *) output_var->name); - return 1; - } - - output_var->type = vt_output; - - err = parse_result_binding(ctx, inst, &output_var->output_binding, Program); - return err; -} - -/** - * This handles variables of the ALIAS kind - * - * \return 0 on sucess, 1 on error - */ -static GLuint -parse_alias (GLcontext * ctx, const GLubyte ** inst, struct var_cache **vc_head, - struct arb_program *Program) -{ - GLuint found; - struct var_cache *temp_var; - - temp_var = parse_string (inst, vc_head, Program, &found); - Program->Position = parse_position (inst); - - if (found) { - program_error2(ctx, Program->Position, - "Duplicate variable declaration", - (char *) temp_var->name); - return 1; - } - - temp_var->type = vt_alias; - temp_var->alias_binding = parse_string (inst, vc_head, Program, &found); - Program->Position = parse_position (inst); - - if (!found) - { - program_error2(ctx, Program->Position, - "Undefined alias value", - (char *) temp_var->alias_binding->name); - return 1; - } - - return 0; -} - -/** - * This handles variables of the ADDRESS kind - * - * \return 0 on sucess, 1 on error - */ -static GLuint -parse_address (GLcontext * ctx, const GLubyte ** inst, struct var_cache **vc_head, - struct arb_program *Program) -{ - GLuint found; - struct var_cache *temp_var; - - while (**inst != 0) { - temp_var = parse_string (inst, vc_head, Program, &found); - Program->Position = parse_position (inst); - if (found) { - program_error2(ctx, Program->Position, - "Duplicate variable declaration", - (char *) temp_var->name); - return 1; - } - - temp_var->type = vt_address; - - if (Program->Base.NumAddressRegs >= - ctx->Const.VertexProgram.MaxAddressRegs) { - const char *msg = "Too many ADDRESS variables declared"; - program_error(ctx, Program->Position, msg); - return 1; - } - - temp_var->address_binding = Program->Base.NumAddressRegs; - Program->Base.NumAddressRegs++; - } - (*inst)++; - - return 0; -} - -/** - * Parse a program declaration - * - * \return 0 on sucess, 1 on error - */ -static GLint -parse_declaration (GLcontext * ctx, const GLubyte ** inst, struct var_cache **vc_head, - struct arb_program *Program) -{ - GLint err = 0; - - switch (*(*inst)++) { - case ADDRESS: - err = parse_address (ctx, inst, vc_head, Program); - break; - - case ALIAS: - err = parse_alias (ctx, inst, vc_head, Program); - break; - - case ATTRIB: - err = parse_attrib (ctx, inst, vc_head, Program); - break; - - case OUTPUT: - err = parse_output (ctx, inst, vc_head, Program); - break; - - case PARAM: - err = parse_param (ctx, inst, vc_head, Program); - break; - - case TEMP: - err = parse_temp (ctx, inst, vc_head, Program); - break; - } - - return err; -} - -/** - * Handle the parsing out of a masked destination register, either for a - * vertex or fragment program. - * - * If we are a vertex program, make sure we don't write to - * result.position if we have specified that the program is - * position invariant - * - * \param File - The register file we write to - * \param Index - The register index we write to - * \param WriteMask - The mask controlling which components we write (1->write) - * - * \return 0 on sucess, 1 on error - */ -static GLuint -parse_masked_dst_reg (GLcontext * ctx, const GLubyte ** inst, - struct var_cache **vc_head, struct arb_program *Program, - gl_register_file *File, GLuint *Index, GLint *WriteMask) -{ - GLuint tmp, result; - struct var_cache *dst; - - /* We either have a result register specified, or a - * variable that may or may not be writable - */ - switch (*(*inst)++) { - case REGISTER_RESULT: - if (parse_result_binding(ctx, inst, Index, Program)) - return 1; - *File = PROGRAM_OUTPUT; - break; - - case REGISTER_ESTABLISHED_NAME: - dst = parse_string (inst, vc_head, Program, &result); - Program->Position = parse_position (inst); - - /* If the name has never been added to our symbol table, we're hosed */ - if (!result) { - program_error(ctx, Program->Position, "0: Undefined variable"); - return 1; - } - - switch (dst->type) { - case vt_output: - *File = PROGRAM_OUTPUT; - *Index = dst->output_binding; - break; - - case vt_temp: - *File = PROGRAM_TEMPORARY; - *Index = dst->temp_binding; - break; - - /* If the var type is not vt_output or vt_temp, no go */ - default: - program_error(ctx, Program->Position, - "Destination register is read only"); - return 1; - } - break; - - default: - program_error(ctx, Program->Position, - "Unexpected opcode in parse_masked_dst_reg()"); - return 1; - } - - - /* Position invariance test */ - /* This test is done now in syntax portion - when position invariance OPTION - is specified, "result.position" rule is disabled so there is no way - to write the position - */ - /*if ((Program->HintPositionInvariant) && (*File == PROGRAM_OUTPUT) && - (*Index == 0)) { - program_error(ctx, Program->Position, - "Vertex program specified position invariance and wrote vertex position"); - }*/ - - /* And then the mask. - * w,a -> bit 0 - * z,b -> bit 1 - * y,g -> bit 2 - * x,r -> bit 3 - * - * ==> Need to reverse the order of bits for this! - */ - tmp = (GLint) *(*inst)++; - *WriteMask = (((tmp>>3) & 0x1) | - ((tmp>>1) & 0x2) | - ((tmp<<1) & 0x4) | - ((tmp<<3) & 0x8)); - - return 0; -} - - -/** - * Handle the parsing of a address register - * - * \param Index - The register index we write to - * - * \return 0 on sucess, 1 on error - */ -static GLuint -parse_address_reg (GLcontext * ctx, const GLubyte ** inst, - struct var_cache **vc_head, - struct arb_program *Program, GLint * Index) -{ - struct var_cache *dst; - GLuint result; - - *Index = 0; /* XXX */ - - dst = parse_string (inst, vc_head, Program, &result); - Program->Position = parse_position (inst); - - /* If the name has never been added to our symbol table, we're hosed */ - if (!result) { - program_error(ctx, Program->Position, "Undefined variable"); - return 1; - } - - if (dst->type != vt_address) { - program_error(ctx, Program->Position, "Variable is not of type ADDRESS"); - return 1; - } - - return 0; -} - -#if 0 /* unused */ -/** - * Handle the parsing out of a masked address register - * - * \param Index - The register index we write to - * \param WriteMask - The mask controlling which components we write (1->write) - * - * \return 0 on sucess, 1 on error - */ -static GLuint -parse_masked_address_reg (GLcontext * ctx, const GLubyte ** inst, - struct var_cache **vc_head, - struct arb_program *Program, GLint * Index, - GLboolean * WriteMask) -{ - if (parse_address_reg (ctx, inst, vc_head, Program, Index)) - return 1; - - /* This should be 0x8 */ - (*inst)++; - - /* Writemask of .x is implied */ - WriteMask[0] = 1; - WriteMask[1] = WriteMask[2] = WriteMask[3] = 0; - - return 0; -} -#endif - -/** - * Parse out a swizzle mask. - * - * Basically convert COMPONENT_X/Y/Z/W to SWIZZLE_X/Y/Z/W - * - * The len parameter allows us to grab 4 components for a vector - * swizzle, or just 1 component for a scalar src register selection - */ -static void -parse_swizzle_mask(const GLubyte ** inst, GLubyte *swizzle, GLint len) -{ - GLint i; - - for (i = 0; i < 4; i++) - swizzle[i] = i; - - for (i = 0; i < len; i++) { - switch (*(*inst)++) { - case COMPONENT_X: - swizzle[i] = SWIZZLE_X; - break; - case COMPONENT_Y: - swizzle[i] = SWIZZLE_Y; - break; - case COMPONENT_Z: - swizzle[i] = SWIZZLE_Z; - break; - case COMPONENT_W: - swizzle[i] = SWIZZLE_W; - break; - default: - _mesa_problem(NULL, "bad component in parse_swizzle_mask()"); - return; - } - } - - if (len == 1) - swizzle[1] = swizzle[2] = swizzle[3] = swizzle[0]; -} - - -/** - * Parse an extended swizzle mask which is a sequence of - * four x/y/z/w/0/1 tokens. - * \return swizzle four swizzle values - * \return negateMask four element bitfield - */ -static void -parse_extended_swizzle_mask(const GLubyte **inst, GLubyte swizzle[4], - GLubyte *negateMask) -{ - GLint i; - - *negateMask = 0x0; - for (i = 0; i < 4; i++) { - GLubyte swz; - if (parse_sign(inst) == -1) - *negateMask |= (1 << i); - - swz = *(*inst)++; - - switch (swz) { - case COMPONENT_0: - swizzle[i] = SWIZZLE_ZERO; - break; - case COMPONENT_1: - swizzle[i] = SWIZZLE_ONE; - break; - case COMPONENT_X: - swizzle[i] = SWIZZLE_X; - break; - case COMPONENT_Y: - swizzle[i] = SWIZZLE_Y; - break; - case COMPONENT_Z: - swizzle[i] = SWIZZLE_Z; - break; - case COMPONENT_W: - swizzle[i] = SWIZZLE_W; - break; - default: - _mesa_problem(NULL, "bad case in parse_extended_swizzle_mask()"); - return; - } - } -} - - -static GLuint -parse_src_reg (GLcontext * ctx, const GLubyte ** inst, - struct var_cache **vc_head, - struct arb_program *Program, - gl_register_file * File, GLint * Index, GLuint *swizzle, - GLboolean *IsRelOffset ) -{ - struct var_cache *src; - GLuint binding, is_generic, found; - GLint offset; - - *IsRelOffset = 0; - - *swizzle = SWIZZLE_XYZW; /* default */ - - /* And the binding for the src */ - switch (*(*inst)++) { - case REGISTER_ATTRIB: - if (parse_attrib_binding - (ctx, inst, Program, &binding, &is_generic)) - return 1; - *File = PROGRAM_INPUT; - *Index = binding; - - /* We need to insert a dummy variable into the var_cache so we can - * catch generic vertex attrib aliasing errors - */ - var_cache_create(&src); - src->type = vt_attrib; - src->name = (const GLubyte *) "Dummy Attrib Variable"; - src->attrib_binding = binding; - src->attrib_is_generic = is_generic; - var_cache_append(vc_head, src); - if (generic_attrib_check(*vc_head)) { - program_error(ctx, Program->Position, - "Cannot use both a generic vertex attribute " - "and a specific attribute of the same type"); - return 1; - } - break; - - case REGISTER_PARAM: - switch (**inst) { - case PARAM_ARRAY_ELEMENT: - (*inst)++; - src = parse_string (inst, vc_head, Program, &found); - Program->Position = parse_position (inst); - - if (!found) { - program_error2(ctx, Program->Position, - "Undefined variable", - (char *) src->name); - return 1; - } - - *File = (gl_register_file) src->param_binding_type; - - switch (*(*inst)++) { - case ARRAY_INDEX_ABSOLUTE: - offset = parse_integer (inst, Program); - - if ((offset < 0) - || (offset >= (int)src->param_binding_length)) { - program_error(ctx, Program->Position, - "Index out of range"); - /* offset, src->name */ - return 1; - } - - *Index = src->param_binding_begin + offset; - *swizzle = src->swizzle; - break; - - case ARRAY_INDEX_RELATIVE: - { - GLint addr_reg_idx, rel_off; - - /* First, grab the address regiseter */ - if (parse_address_reg (ctx, inst, vc_head, Program, &addr_reg_idx)) - return 1; - - /* And the .x */ - ((*inst)++); - ((*inst)++); - ((*inst)++); - ((*inst)++); - - /* Then the relative offset */ - if (parse_relative_offset(ctx, inst, Program, &rel_off)) return 1; - - /* And store it properly */ - *Index = src->param_binding_begin + rel_off; - *IsRelOffset = 1; - *swizzle = src->swizzle; - } - break; - } - break; - - default: - if (parse_param_use (ctx, inst, vc_head, Program, &src)) - return 1; - - *File = (gl_register_file) src->param_binding_type; - *Index = src->param_binding_begin; - *swizzle = src->swizzle; - break; - } - break; - - case REGISTER_ESTABLISHED_NAME: - src = parse_string (inst, vc_head, Program, &found); - Program->Position = parse_position (inst); - - /* If the name has never been added to our symbol table, we're hosed */ - if (!found) { - program_error(ctx, Program->Position, - "3: Undefined variable"); /* src->name */ - return 1; - } - - switch (src->type) { - case vt_attrib: - *File = PROGRAM_INPUT; - *Index = src->attrib_binding; - break; - - /* XXX: We have to handle offsets someplace in here! -- or are those above? */ - case vt_param: - *File = (gl_register_file) src->param_binding_type; - *Index = src->param_binding_begin; - break; - - case vt_temp: - *File = PROGRAM_TEMPORARY; - *Index = src->temp_binding; - break; - - /* If the var type is vt_output no go */ - default: - program_error(ctx, Program->Position, - "destination register is read only"); - /* bad src->name */ - return 1; - } - break; - - default: - program_error(ctx, Program->Position, - "Unknown token in parse_src_reg"); - return 1; - } - - if (*File == PROGRAM_STATE_VAR) { - gl_register_file file; - - /* If we're referencing the Program->Parameters[] array, check if the - * parameter is really a constant/literal. If so, set File to CONSTANT. - */ - assert(*Index < (GLint) Program->Base.Parameters->NumParameters); - file = Program->Base.Parameters->Parameters[*Index].Type; - if (file == PROGRAM_CONSTANT) - *File = PROGRAM_CONSTANT; - } - - /* Add attributes to InputsRead only if they are used the program. - * This avoids the handling of unused ATTRIB declarations in the drivers. */ - if (*File == PROGRAM_INPUT) - Program->Base.InputsRead |= (1 << *Index); - - return 0; -} - - -static GLuint -swizzle_swizzle(GLuint baseSwizzle, const GLubyte swizzle[4]) -{ - GLuint i, swz, s[4]; - for (i = 0; i < 4; i++) { - GLuint c = swizzle[i]; - if (c <= SWIZZLE_W) - s[i] = GET_SWZ(baseSwizzle, c); - else - s[i] = c; - } - swz = MAKE_SWIZZLE4(s[0], s[1], s[2], s[3]); - return swz; -} - -/** - * Parse vertex/fragment program vector source register. - */ -static GLuint -parse_vector_src_reg(GLcontext *ctx, const GLubyte **inst, - struct var_cache **vc_head, - struct arb_program *program, - struct prog_src_register *reg) -{ - gl_register_file file; - GLint index; - GLubyte negateMask; - GLubyte swizzle[4]; - GLboolean isRelOffset; - GLuint baseSwizzle; - - /* Grab the sign */ - negateMask = (parse_sign (inst) == -1) ? NEGATE_XYZW : NEGATE_NONE; - - /* And the src reg */ - if (parse_src_reg(ctx, inst, vc_head, program, &file, &index, &baseSwizzle, - &isRelOffset)) - return 1; - - /* finally, the swizzle */ - parse_swizzle_mask(inst, swizzle, 4); - - reg->File = file; - reg->Index = index; - reg->Swizzle = swizzle_swizzle(baseSwizzle, swizzle); - reg->Negate = negateMask; - reg->RelAddr = isRelOffset; - return 0; -} - - -/** - * Parse vertex/fragment program scalar source register. - */ -static GLuint -parse_scalar_src_reg(GLcontext *ctx, const GLubyte **inst, - struct var_cache **vc_head, - struct arb_program *program, - struct prog_src_register *reg) -{ - gl_register_file file; - GLint index; - GLubyte negateMask; - GLubyte swizzle[4]; - GLboolean isRelOffset; - GLuint baseSwizzle; - - /* Grab the sign */ - negateMask = (parse_sign (inst) == -1) ? NEGATE_XYZW : NEGATE_NONE; - - /* And the src reg */ - if (parse_src_reg(ctx, inst, vc_head, program, &file, &index, &baseSwizzle, - &isRelOffset)) - return 1; - - /* finally, the swizzle */ - parse_swizzle_mask(inst, swizzle, 1); - - reg->File = file; - reg->Index = index; - reg->Swizzle = swizzle_swizzle(baseSwizzle, swizzle); - reg->Negate = negateMask; - reg->RelAddr = isRelOffset; - return 0; -} - - -/** - * Parse vertex/fragment program destination register. - * \return 1 if error, 0 if no error. - */ -static GLuint -parse_dst_reg(GLcontext * ctx, const GLubyte ** inst, - struct var_cache **vc_head, struct arb_program *program, - struct prog_dst_register *reg ) -{ - GLint mask; - GLuint idx; - gl_register_file file; - - if (parse_masked_dst_reg (ctx, inst, vc_head, program, &file, &idx, &mask)) - return 1; - - reg->File = file; - reg->Index = idx; - reg->WriteMask = mask; - return 0; -} - - -/** - * This is a big mother that handles getting opcodes into the instruction - * and handling the src & dst registers for fragment program instructions - * \return 1 if error, 0 if no error - */ -static GLuint -parse_fp_instruction (GLcontext * ctx, const GLubyte ** inst, - struct var_cache **vc_head, struct arb_program *Program, - struct prog_instruction *fp) -{ - GLint a; - GLuint texcoord; - GLubyte instClass, type, code; - GLboolean rel; - GLuint shadow_tex = 0; - - _mesa_init_instructions(fp, 1); - - /* OP_ALU_INST or OP_TEX_INST */ - instClass = *(*inst)++; - - /* OP_ALU_{VECTOR, SCALAR, BINSC, BIN, TRI, SWZ}, - * OP_TEX_{SAMPLE, KIL} - */ - type = *(*inst)++; - - /* The actual opcode name */ - code = *(*inst)++; - - /* Increment the correct count */ - switch (instClass) { - case OP_ALU_INST: - Program->NumAluInstructions++; - break; - case OP_TEX_INST: - Program->NumTexInstructions++; - break; - } - - switch (type) { - case OP_ALU_VECTOR: - switch (code) { - case OP_ABS_SAT: - fp->SaturateMode = SATURATE_ZERO_ONE; - case OP_ABS: - fp->Opcode = OPCODE_ABS; - break; - - case OP_FLR_SAT: - fp->SaturateMode = SATURATE_ZERO_ONE; - case OP_FLR: - fp->Opcode = OPCODE_FLR; - break; - - case OP_FRC_SAT: - fp->SaturateMode = SATURATE_ZERO_ONE; - case OP_FRC: - fp->Opcode = OPCODE_FRC; - break; - - case OP_LIT_SAT: - fp->SaturateMode = SATURATE_ZERO_ONE; - case OP_LIT: - fp->Opcode = OPCODE_LIT; - break; - - case OP_MOV_SAT: - fp->SaturateMode = SATURATE_ZERO_ONE; - case OP_MOV: - fp->Opcode = OPCODE_MOV; - break; - } - - if (parse_dst_reg (ctx, inst, vc_head, Program, &fp->DstReg)) - return 1; - - if (parse_vector_src_reg(ctx, inst, vc_head, Program, &fp->SrcReg[0])) - return 1; - break; - - case OP_ALU_SCALAR: - switch (code) { - case OP_COS_SAT: - fp->SaturateMode = SATURATE_ZERO_ONE; - case OP_COS: - fp->Opcode = OPCODE_COS; - break; - - case OP_EX2_SAT: - fp->SaturateMode = SATURATE_ZERO_ONE; - case OP_EX2: - fp->Opcode = OPCODE_EX2; - break; - - case OP_LG2_SAT: - fp->SaturateMode = SATURATE_ZERO_ONE; - case OP_LG2: - fp->Opcode = OPCODE_LG2; - break; - - case OP_RCP_SAT: - fp->SaturateMode = SATURATE_ZERO_ONE; - case OP_RCP: - fp->Opcode = OPCODE_RCP; - break; - - case OP_RSQ_SAT: - fp->SaturateMode = SATURATE_ZERO_ONE; - case OP_RSQ: - fp->Opcode = OPCODE_RSQ; - break; - - case OP_SIN_SAT: - fp->SaturateMode = SATURATE_ZERO_ONE; - case OP_SIN: - fp->Opcode = OPCODE_SIN; - break; - - case OP_SCS_SAT: - fp->SaturateMode = SATURATE_ZERO_ONE; - case OP_SCS: - - fp->Opcode = OPCODE_SCS; - break; - } - - if (parse_dst_reg (ctx, inst, vc_head, Program, &fp->DstReg)) - return 1; - - if (parse_scalar_src_reg(ctx, inst, vc_head, Program, &fp->SrcReg[0])) - return 1; - break; - - case OP_ALU_BINSC: - switch (code) { - case OP_POW_SAT: - fp->SaturateMode = SATURATE_ZERO_ONE; - case OP_POW: - fp->Opcode = OPCODE_POW; - break; - } - - if (parse_dst_reg(ctx, inst, vc_head, Program, &fp->DstReg)) - return 1; - - for (a = 0; a < 2; a++) { - if (parse_scalar_src_reg(ctx, inst, vc_head, Program, &fp->SrcReg[a])) - return 1; - } - break; - - - case OP_ALU_BIN: - switch (code) { - case OP_ADD_SAT: - fp->SaturateMode = SATURATE_ZERO_ONE; - case OP_ADD: - fp->Opcode = OPCODE_ADD; - break; - - case OP_DP3_SAT: - fp->SaturateMode = SATURATE_ZERO_ONE; - case OP_DP3: - fp->Opcode = OPCODE_DP3; - break; - - case OP_DP4_SAT: - fp->SaturateMode = SATURATE_ZERO_ONE; - case OP_DP4: - fp->Opcode = OPCODE_DP4; - break; - - case OP_DPH_SAT: - fp->SaturateMode = SATURATE_ZERO_ONE; - case OP_DPH: - fp->Opcode = OPCODE_DPH; - break; - - case OP_DST_SAT: - fp->SaturateMode = SATURATE_ZERO_ONE; - case OP_DST: - fp->Opcode = OPCODE_DST; - break; - - case OP_MAX_SAT: - fp->SaturateMode = SATURATE_ZERO_ONE; - case OP_MAX: - fp->Opcode = OPCODE_MAX; - break; - - case OP_MIN_SAT: - fp->SaturateMode = SATURATE_ZERO_ONE; - case OP_MIN: - fp->Opcode = OPCODE_MIN; - break; - - case OP_MUL_SAT: - fp->SaturateMode = SATURATE_ZERO_ONE; - case OP_MUL: - fp->Opcode = OPCODE_MUL; - break; - - case OP_SGE_SAT: - fp->SaturateMode = SATURATE_ZERO_ONE; - case OP_SGE: - fp->Opcode = OPCODE_SGE; - break; - - case OP_SLT_SAT: - fp->SaturateMode = SATURATE_ZERO_ONE; - case OP_SLT: - fp->Opcode = OPCODE_SLT; - break; - - case OP_SUB_SAT: - fp->SaturateMode = SATURATE_ZERO_ONE; - case OP_SUB: - fp->Opcode = OPCODE_SUB; - break; - - case OP_XPD_SAT: - fp->SaturateMode = SATURATE_ZERO_ONE; - case OP_XPD: - fp->Opcode = OPCODE_XPD; - break; - } - - if (parse_dst_reg (ctx, inst, vc_head, Program, &fp->DstReg)) - return 1; - for (a = 0; a < 2; a++) { - if (parse_vector_src_reg(ctx, inst, vc_head, Program, &fp->SrcReg[a])) - return 1; - } - break; - - case OP_ALU_TRI: - switch (code) { - case OP_CMP_SAT: - fp->SaturateMode = SATURATE_ZERO_ONE; - case OP_CMP: - fp->Opcode = OPCODE_CMP; - break; - - case OP_LRP_SAT: - fp->SaturateMode = SATURATE_ZERO_ONE; - case OP_LRP: - fp->Opcode = OPCODE_LRP; - break; - - case OP_MAD_SAT: - fp->SaturateMode = SATURATE_ZERO_ONE; - case OP_MAD: - fp->Opcode = OPCODE_MAD; - break; - } - - if (parse_dst_reg (ctx, inst, vc_head, Program, &fp->DstReg)) - return 1; - - for (a = 0; a < 3; a++) { - if (parse_vector_src_reg(ctx, inst, vc_head, Program, &fp->SrcReg[a])) - return 1; - } - break; - - case OP_ALU_SWZ: - switch (code) { - case OP_SWZ_SAT: - fp->SaturateMode = SATURATE_ZERO_ONE; - case OP_SWZ: - fp->Opcode = OPCODE_SWZ; - break; - } - if (parse_dst_reg (ctx, inst, vc_head, Program, &fp->DstReg)) - return 1; - - { - GLubyte swizzle[4]; - GLubyte negateMask; - gl_register_file file; - GLint index; - GLuint baseSwizzle; - - if (parse_src_reg(ctx, inst, vc_head, Program, &file, &index, - &baseSwizzle, &rel)) - return 1; - parse_extended_swizzle_mask(inst, swizzle, &negateMask); - fp->SrcReg[0].File = file; - fp->SrcReg[0].Index = index; - fp->SrcReg[0].Negate = negateMask; - fp->SrcReg[0].Swizzle = MAKE_SWIZZLE4(swizzle[0], - swizzle[1], - swizzle[2], - swizzle[3]); - } - break; - - case OP_TEX_SAMPLE: - switch (code) { - case OP_TEX_SAT: - fp->SaturateMode = SATURATE_ZERO_ONE; - case OP_TEX: - fp->Opcode = OPCODE_TEX; - break; - - case OP_TXP_SAT: - fp->SaturateMode = SATURATE_ZERO_ONE; - case OP_TXP: - fp->Opcode = OPCODE_TXP; - break; - - case OP_TXB_SAT: - fp->SaturateMode = SATURATE_ZERO_ONE; - case OP_TXB: - fp->Opcode = OPCODE_TXB; - break; - } - - if (parse_dst_reg (ctx, inst, vc_head, Program, &fp->DstReg)) - return 1; - - if (parse_vector_src_reg(ctx, inst, vc_head, Program, &fp->SrcReg[0])) - return 1; - - /* texImageUnit */ - if (parse_teximage_num (ctx, inst, Program, &texcoord)) - return 1; - fp->TexSrcUnit = texcoord; - - /* texTarget */ - switch (*(*inst)++) { - case TEXTARGET_SHADOW1D: - shadow_tex = 1 << texcoord; - /* FALLTHROUGH */ - case TEXTARGET_1D: - fp->TexSrcTarget = TEXTURE_1D_INDEX; - break; - case TEXTARGET_SHADOW2D: - shadow_tex = 1 << texcoord; - /* FALLTHROUGH */ - case TEXTARGET_2D: - fp->TexSrcTarget = TEXTURE_2D_INDEX; - break; - case TEXTARGET_3D: - fp->TexSrcTarget = TEXTURE_3D_INDEX; - break; - case TEXTARGET_SHADOWRECT: - shadow_tex = 1 << texcoord; - /* FALLTHROUGH */ - case TEXTARGET_RECT: - fp->TexSrcTarget = TEXTURE_RECT_INDEX; - break; - case TEXTARGET_CUBE: - fp->TexSrcTarget = TEXTURE_CUBE_INDEX; - break; - case TEXTARGET_SHADOW1D_ARRAY: - shadow_tex = 1 << texcoord; - /* FALLTHROUGH */ - case TEXTARGET_1D_ARRAY: - fp->TexSrcTarget = TEXTURE_1D_ARRAY_INDEX; - break; - case TEXTARGET_SHADOW2D_ARRAY: - shadow_tex = 1 << texcoord; - /* FALLTHROUGH */ - case TEXTARGET_2D_ARRAY: - fp->TexSrcTarget = TEXTURE_2D_ARRAY_INDEX; - break; - } - - if (shadow_tex) - fp->TexShadow = 1; - - /* Don't test the first time a particular sampler is seen. Each time - * after that, make sure the shadow state is the same. - */ - if ((_mesa_bitcount(Program->TexturesUsed[texcoord]) > 0) - && ((Program->ShadowSamplers & (1 << texcoord)) != shadow_tex)) { - program_error(ctx, Program->Position, - "texture image unit used for shadow sampling and non-shadow sampling"); - return 1; - } - - Program->TexturesUsed[texcoord] |= (1 << fp->TexSrcTarget); - /* Check that both "2D" and "CUBE" (for example) aren't both used */ - if (_mesa_bitcount(Program->TexturesUsed[texcoord]) > 1) { - program_error(ctx, Program->Position, - "multiple targets used on one texture image unit"); - return 1; - } - - - Program->ShadowSamplers |= shadow_tex; - break; - - case OP_TEX_KIL: - Program->UsesKill = 1; - if (parse_vector_src_reg(ctx, inst, vc_head, Program, &fp->SrcReg[0])) - return 1; - fp->Opcode = OPCODE_KIL; - break; - default: - _mesa_problem(ctx, "bad type 0x%x in parse_fp_instruction()", type); - return 1; - } - - return 0; -} - - -/** - * Handle the parsing out of a masked address register - * - * \param Index - The register index we write to - * \param WriteMask - The mask controlling which components we write (1->write) - * - * \return 0 on sucess, 1 on error - */ -static GLuint -parse_vp_address_reg (GLcontext * ctx, const GLubyte ** inst, - struct var_cache **vc_head, - struct arb_program *Program, - struct prog_dst_register *reg) -{ - GLint idx; - - if (parse_address_reg (ctx, inst, vc_head, Program, &idx)) - return 1; - - /* This should be 0x8 */ - (*inst)++; - - reg->File = PROGRAM_ADDRESS; - reg->Index = idx; - - /* Writemask of .x is implied */ - reg->WriteMask = 0x1; - return 0; -} - - -/** - * This is a big mother that handles getting opcodes into the instruction - * and handling the src & dst registers for vertex program instructions - */ -static GLuint -parse_vp_instruction (GLcontext * ctx, const GLubyte ** inst, - struct var_cache **vc_head, struct arb_program *Program, - struct prog_instruction *vp) -{ - GLint a; - GLubyte type, code; - - /* OP_ALU_{ARL, VECTOR, SCALAR, BINSC, BIN, TRI, SWZ} */ - type = *(*inst)++; - - /* The actual opcode name */ - code = *(*inst)++; - - _mesa_init_instructions(vp, 1); - - switch (type) { - /* XXX: */ - case OP_ALU_ARL: - vp->Opcode = OPCODE_ARL; - - /* Remember to set SrcReg.RelAddr; */ - - /* Get the masked address register [dst] */ - if (parse_vp_address_reg(ctx, inst, vc_head, Program, &vp->DstReg)) - return 1; - - vp->DstReg.File = PROGRAM_ADDRESS; - - /* Get a scalar src register */ - if (parse_scalar_src_reg(ctx, inst, vc_head, Program, &vp->SrcReg[0])) - return 1; - - break; - - case OP_ALU_VECTOR: - switch (code) { - case OP_ABS: - vp->Opcode = OPCODE_ABS; - break; - case OP_FLR: - vp->Opcode = OPCODE_FLR; - break; - case OP_FRC: - vp->Opcode = OPCODE_FRC; - break; - case OP_LIT: - vp->Opcode = OPCODE_LIT; - break; - case OP_MOV: - vp->Opcode = OPCODE_MOV; - break; - } - - if (parse_dst_reg(ctx, inst, vc_head, Program, &vp->DstReg)) - return 1; - - if (parse_vector_src_reg(ctx, inst, vc_head, Program, &vp->SrcReg[0])) - return 1; - break; - - case OP_ALU_SCALAR: - switch (code) { - case OP_EX2: - vp->Opcode = OPCODE_EX2; - break; - case OP_EXP: - vp->Opcode = OPCODE_EXP; - break; - case OP_LG2: - vp->Opcode = OPCODE_LG2; - break; - case OP_LOG: - vp->Opcode = OPCODE_LOG; - break; - case OP_RCP: - vp->Opcode = OPCODE_RCP; - break; - case OP_RSQ: - vp->Opcode = OPCODE_RSQ; - break; - } - if (parse_dst_reg(ctx, inst, vc_head, Program, &vp->DstReg)) - return 1; - - if (parse_scalar_src_reg(ctx, inst, vc_head, Program, &vp->SrcReg[0])) - return 1; - break; - - case OP_ALU_BINSC: - switch (code) { - case OP_POW: - vp->Opcode = OPCODE_POW; - break; - } - if (parse_dst_reg(ctx, inst, vc_head, Program, &vp->DstReg)) - return 1; - - for (a = 0; a < 2; a++) { - if (parse_scalar_src_reg(ctx, inst, vc_head, Program, &vp->SrcReg[a])) - return 1; - } - break; - - case OP_ALU_BIN: - switch (code) { - case OP_ADD: - vp->Opcode = OPCODE_ADD; - break; - case OP_DP3: - vp->Opcode = OPCODE_DP3; - break; - case OP_DP4: - vp->Opcode = OPCODE_DP4; - break; - case OP_DPH: - vp->Opcode = OPCODE_DPH; - break; - case OP_DST: - vp->Opcode = OPCODE_DST; - break; - case OP_MAX: - vp->Opcode = OPCODE_MAX; - break; - case OP_MIN: - vp->Opcode = OPCODE_MIN; - break; - case OP_MUL: - vp->Opcode = OPCODE_MUL; - break; - case OP_SGE: - vp->Opcode = OPCODE_SGE; - break; - case OP_SLT: - vp->Opcode = OPCODE_SLT; - break; - case OP_SUB: - vp->Opcode = OPCODE_SUB; - break; - case OP_XPD: - vp->Opcode = OPCODE_XPD; - break; - } - if (parse_dst_reg(ctx, inst, vc_head, Program, &vp->DstReg)) - return 1; - - for (a = 0; a < 2; a++) { - if (parse_vector_src_reg(ctx, inst, vc_head, Program, &vp->SrcReg[a])) - return 1; - } - break; - - case OP_ALU_TRI: - switch (code) { - case OP_MAD: - vp->Opcode = OPCODE_MAD; - break; - } - - if (parse_dst_reg(ctx, inst, vc_head, Program, &vp->DstReg)) - return 1; - - for (a = 0; a < 3; a++) { - if (parse_vector_src_reg(ctx, inst, vc_head, Program, &vp->SrcReg[a])) - return 1; - } - break; - - case OP_ALU_SWZ: - switch (code) { - case OP_SWZ: - vp->Opcode = OPCODE_SWZ; - break; - } - { - GLubyte swizzle[4]; - GLubyte negateMask; - GLboolean relAddr; - gl_register_file file; - GLint index; - GLuint baseSwizzle; - - if (parse_dst_reg(ctx, inst, vc_head, Program, &vp->DstReg)) - return 1; - - if (parse_src_reg(ctx, inst, vc_head, Program, &file, &index, - &baseSwizzle, &relAddr)) - return 1; - parse_extended_swizzle_mask (inst, swizzle, &negateMask); - vp->SrcReg[0].File = file; - vp->SrcReg[0].Index = index; - vp->SrcReg[0].Negate = negateMask; - vp->SrcReg[0].Swizzle = MAKE_SWIZZLE4(swizzle[0], - swizzle[1], - swizzle[2], - swizzle[3]); - vp->SrcReg[0].RelAddr = relAddr; - } - break; - } - return 0; -} - -#if DEBUG_PARSING - -static GLvoid -debug_variables (GLcontext * ctx, struct var_cache *vc_head, - struct arb_program *Program) -{ - struct var_cache *vc; - GLint a, b; - - fprintf (stderr, "debug_variables, vc_head: %p\n", (void*) vc_head); - - /* First of all, print out the contents of the var_cache */ - vc = vc_head; - while (vc) { - fprintf (stderr, "[%p]\n", (void*) vc); - switch (vc->type) { - case vt_none: - fprintf (stderr, "UNDEFINED %s\n", vc->name); - break; - case vt_attrib: - fprintf (stderr, "ATTRIB %s\n", vc->name); - fprintf (stderr, " binding: 0x%x\n", vc->attrib_binding); - break; - case vt_param: - fprintf (stderr, "PARAM %s begin: %d len: %d\n", vc->name, - vc->param_binding_begin, vc->param_binding_length); - b = vc->param_binding_begin; - for (a = 0; a < vc->param_binding_length; a++) { - fprintf (stderr, "%s\n", - Program->Base.Parameters->Parameters[a + b].Name); - if (Program->Base.Parameters->Parameters[a + b].Type == PROGRAM_STATE_VAR) { - char *s; - s = _mesa_program_state_string(Program->Base.Parameters->Parameters - [a + b].StateIndexes); - fprintf(stderr, "%s\n", s); - _mesa_free(s); - } - else - fprintf (stderr, "%f %f %f %f\n", - Program->Base.Parameters->ParameterValues[a + b][0], - Program->Base.Parameters->ParameterValues[a + b][1], - Program->Base.Parameters->ParameterValues[a + b][2], - Program->Base.Parameters->ParameterValues[a + b][3]); - } - break; - case vt_temp: - fprintf (stderr, "TEMP %s\n", vc->name); - fprintf (stderr, " binding: 0x%x\n", vc->temp_binding); - break; - case vt_output: - fprintf (stderr, "OUTPUT %s\n", vc->name); - fprintf (stderr, " binding: 0x%x\n", vc->output_binding); - break; - case vt_alias: - fprintf (stderr, "ALIAS %s\n", vc->name); - fprintf (stderr, " binding: 0x%p (%s)\n", - (void*) vc->alias_binding, vc->alias_binding->name); - break; - default: - /* nothing */ - ; - } - vc = vc->next; - } -} - -#endif /* DEBUG_PARSING */ - - -/** - * The main loop for parsing a fragment or vertex program - * - * \return 1 on error, 0 on success - */ -static GLint -parse_instructions(GLcontext * ctx, const GLubyte * inst, - struct var_cache **vc_head, struct arb_program *Program) -{ - const GLuint maxInst = (Program->Base.Target == GL_FRAGMENT_PROGRAM_ARB) - ? ctx->Const.FragmentProgram.MaxInstructions - : ctx->Const.VertexProgram.MaxInstructions; - GLint err = 0; - - ASSERT(MAX_PROGRAM_INSTRUCTIONS >= maxInst); - - Program->MajorVersion = (GLuint) * inst++; - Program->MinorVersion = (GLuint) * inst++; - - while (*inst != END) { - switch (*inst++) { - - case OPTION: - switch (*inst++) { - case ARB_PRECISION_HINT_FASTEST: - Program->PrecisionOption = GL_FASTEST; - break; - - case ARB_PRECISION_HINT_NICEST: - Program->PrecisionOption = GL_NICEST; - break; - - case ARB_FOG_EXP: - Program->FogOption = GL_EXP; - break; - - case ARB_FOG_EXP2: - Program->FogOption = GL_EXP2; - break; - - case ARB_FOG_LINEAR: - Program->FogOption = GL_LINEAR; - break; - - case ARB_POSITION_INVARIANT: - if (Program->Base.Target == GL_VERTEX_PROGRAM_ARB) - Program->HintPositionInvariant = GL_TRUE; - break; - - case ARB_FRAGMENT_PROGRAM_SHADOW: - if (Program->Base.Target == GL_FRAGMENT_PROGRAM_ARB) { - /* TODO ARB_fragment_program_shadow code */ - } - break; - - case ARB_DRAW_BUFFERS: - if (Program->Base.Target == GL_FRAGMENT_PROGRAM_ARB) { - /* do nothing for now */ - } - break; - - case MESA_TEXTURE_ARRAY: - /* do nothing for now */ - break; - } - break; - - case INSTRUCTION: - /* check length */ - if (Program->Base.NumInstructions + 1 >= maxInst) { - program_error(ctx, Program->Position, - "Max instruction count exceeded"); - return 1; - } - Program->Position = parse_position (&inst); - /* parse the current instruction */ - if (Program->Base.Target == GL_FRAGMENT_PROGRAM_ARB) { - err = parse_fp_instruction (ctx, &inst, vc_head, Program, - &Program->Base.Instructions[Program->Base.NumInstructions]); - } - else { - err = parse_vp_instruction (ctx, &inst, vc_head, Program, - &Program->Base.Instructions[Program->Base.NumInstructions]); - } - - /* increment instuction count */ - Program->Base.NumInstructions++; - break; - - case DECLARATION: - err = parse_declaration (ctx, &inst, vc_head, Program); - break; - - default: - break; - } - - if (err) - break; - } - - /* Finally, tag on an OPCODE_END instruction */ - { - const GLuint numInst = Program->Base.NumInstructions; - _mesa_init_instructions(Program->Base.Instructions + numInst, 1); - Program->Base.Instructions[numInst].Opcode = OPCODE_END; - } - Program->Base.NumInstructions++; - - /* - * Initialize native counts to logical counts. The device driver may - * change them if program is translated into a hardware program. - */ - Program->Base.NumNativeInstructions = Program->Base.NumInstructions; - Program->Base.NumNativeTemporaries = Program->Base.NumTemporaries; - Program->Base.NumNativeParameters = Program->Base.NumParameters; - Program->Base.NumNativeAttributes = Program->Base.NumAttributes; - Program->Base.NumNativeAddressRegs = Program->Base.NumAddressRegs; - - return err; -} - - -/* XXX temporary */ -LONGSTRING static char core_grammar_text[] = -#include "shader/grammar/grammar_syn.h" -; - - -/** - * Set a grammar parameter. - * \param name the grammar parameter - * \param value the new parameter value - * \return 0 if OK, 1 if error - */ -static int -set_reg8 (GLcontext *ctx, grammar id, const char *name, GLubyte value) -{ - char error_msg[300]; - GLint error_pos; - - if (grammar_set_reg8 (id, (const byte *) name, value)) - return 0; - - grammar_get_last_error ((byte *) error_msg, 300, &error_pos); - _mesa_set_program_error (ctx, error_pos, error_msg); - _mesa_error (ctx, GL_INVALID_OPERATION, "Grammar Register Error"); - return 1; -} - - -/** - * Enable support for the given language option in the parser. - * \return 1 if OK, 0 if error - */ -static int -enable_ext(GLcontext *ctx, grammar id, const char *name) -{ - return !set_reg8(ctx, id, name, 1); -} - - -/** - * Enable parser extensions based on which OpenGL extensions are supported - * by this rendering context. - * - * \return GL_TRUE if OK, GL_FALSE if error. - */ -static GLboolean -enable_parser_extensions(GLcontext *ctx, grammar id) -{ -#if 0 - /* These are not supported at this time */ - if ((ctx->Extensions.ARB_vertex_blend || - ctx->Extensions.EXT_vertex_weighting) - && !enable_ext(ctx, id, "vertex_blend")) - return GL_FALSE; - if (ctx->Extensions.ARB_matrix_palette - && !enable_ext(ctx, id, "matrix_palette")) - return GL_FALSE; -#endif - if (ctx->Extensions.ARB_fragment_program_shadow - && !enable_ext(ctx, id, "fragment_program_shadow")) - return GL_FALSE; - if (ctx->Extensions.EXT_point_parameters - && !enable_ext(ctx, id, "point_parameters")) - return GL_FALSE; - if (ctx->Extensions.EXT_secondary_color - && !enable_ext(ctx, id, "secondary_color")) - return GL_FALSE; - if (ctx->Extensions.EXT_fog_coord - && !enable_ext(ctx, id, "fog_coord")) - return GL_FALSE; - if (ctx->Extensions.NV_texture_rectangle - && !enable_ext(ctx, id, "texture_rectangle")) - return GL_FALSE; - if (!enable_ext(ctx, id, "draw_buffers")) - return GL_FALSE; - if (ctx->Extensions.MESA_texture_array - && !enable_ext(ctx, id, "texture_array")) - return GL_FALSE; -#if 1 - /* hack for Warcraft (see bug 8060) */ - enable_ext(ctx, id, "vertex_blend"); -#endif - - return GL_TRUE; -} - - -/** - * This kicks everything off. - * - * \param ctx - The GL Context - * \param str - The program string - * \param len - The program string length - * \param program - The arb_program struct to return all the parsed info in - * \return GL_TRUE on sucess, GL_FALSE on error - */ -static GLboolean -_mesa_parse_arb_program(GLcontext *ctx, GLenum target, - const GLubyte *str, GLsizei len, - struct arb_program *program) -{ - GLint a, err, error_pos; - char error_msg[300]; - GLuint parsed_len; - struct var_cache *vc_head; - grammar arbprogram_syn_id; - GLubyte *parsed, *inst; - GLubyte *strz = NULL; - static int arbprogram_syn_is_ok = 0; /* XXX temporary */ - - /* set the program target before parsing */ - program->Base.Target = target; - - /* Reset error state */ - _mesa_set_program_error(ctx, -1, NULL); - - /* check if arb_grammar_text (arbprogram.syn) is syntactically correct */ - if (!arbprogram_syn_is_ok) { - /* One-time initialization of parsing system */ - grammar grammar_syn_id; - GLuint parsed_len; - - grammar_syn_id = grammar_load_from_text ((byte *) core_grammar_text); - if (grammar_syn_id == 0) { - grammar_get_last_error ((byte *) error_msg, 300, &error_pos); - /* XXX this is not a GL error - it's an implementation bug! - FIX */ - _mesa_set_program_error (ctx, error_pos, error_msg); - _mesa_error (ctx, GL_INVALID_OPERATION, - "glProgramStringARB(Error loading grammar rule set)"); - return GL_FALSE; - } - - err = !grammar_check(grammar_syn_id, (byte *) arb_grammar_text, - &parsed, &parsed_len); - - /* 'parsed' is unused here */ - _mesa_free (parsed); - parsed = NULL; - - /* NOTE: we can't destroy grammar_syn_id right here because - * grammar_destroy() can reset the last error - */ - if (err) { - /* XXX this is not a GL error - it's an implementation bug! - FIX */ - grammar_get_last_error ((byte *) error_msg, 300, &error_pos); - _mesa_set_program_error (ctx, error_pos, error_msg); - _mesa_error (ctx, GL_INVALID_OPERATION, - "glProgramString(Error loading grammar rule set"); - grammar_destroy (grammar_syn_id); - return GL_FALSE; - } - - grammar_destroy (grammar_syn_id); - - arbprogram_syn_is_ok = 1; - } - - /* create the grammar object */ - arbprogram_syn_id = grammar_load_from_text ((byte *) arb_grammar_text); - if (arbprogram_syn_id == 0) { - /* XXX this is not a GL error - it's an implementation bug! - FIX */ - grammar_get_last_error ((GLubyte *) error_msg, 300, &error_pos); - _mesa_set_program_error (ctx, error_pos, error_msg); - _mesa_error (ctx, GL_INVALID_OPERATION, - "glProgramString(Error loading grammer rule set)"); - return GL_FALSE; - } - - /* Set program_target register value */ - if (set_reg8 (ctx, arbprogram_syn_id, "program_target", - program->Base.Target == GL_FRAGMENT_PROGRAM_ARB ? 0x10 : 0x20)) { - grammar_destroy (arbprogram_syn_id); - return GL_FALSE; - } - - if (!enable_parser_extensions(ctx, arbprogram_syn_id)) { - grammar_destroy(arbprogram_syn_id); - return GL_FALSE; - } - - /* check for NULL character occurences */ - { - GLint i; - for (i = 0; i < len; i++) { - if (str[i] == '\0') { - program_error(ctx, i, "illegal character"); - grammar_destroy (arbprogram_syn_id); - return GL_FALSE; - } - } - } - - /* copy the program string to a null-terminated string */ - strz = (GLubyte *) _mesa_malloc (len + 1); - if (!strz) { - _mesa_error(ctx, GL_OUT_OF_MEMORY, "glProgramStringARB"); - grammar_destroy (arbprogram_syn_id); - return GL_FALSE; - } - _mesa_memcpy (strz, str, len); - strz[len] = '\0'; - - /* do a fast check on program string - initial production buffer is 4K */ - err = !grammar_fast_check(arbprogram_syn_id, strz, - &parsed, &parsed_len, 0x1000); - - /* Syntax parse error */ - if (err) { - grammar_get_last_error((GLubyte *) error_msg, 300, &error_pos); - program_error(ctx, error_pos, error_msg); - -#if DEBUG_PARSING - /* useful for debugging */ - do { - int line, col; - char *s; - fprintf(stderr, "program: %s\n", (char *) strz); - fprintf(stderr, "Error Pos: %d\n", ctx->Program.ErrorPos); - s = (char *) _mesa_find_line_column(strz, strz+ctx->Program.ErrorPos, - &line, &col); - fprintf(stderr, "line %d col %d: %s\n", line, col, s); - } while (0); -#endif - - _mesa_free(strz); - _mesa_free(parsed); - - grammar_destroy (arbprogram_syn_id); - return GL_FALSE; - } - - grammar_destroy (arbprogram_syn_id); - - /* - * Program string is syntactically correct at this point - * Parse the tokenized version of the program now, generating - * vertex/fragment program instructions. - */ - - /* Initialize the arb_program struct */ - program->Base.String = strz; - program->Base.Instructions = _mesa_alloc_instructions(MAX_PROGRAM_INSTRUCTIONS); - program->Base.NumInstructions = - program->Base.NumTemporaries = - program->Base.NumParameters = - program->Base.NumAttributes = program->Base.NumAddressRegs = 0; - program->Base.Parameters = _mesa_new_parameter_list (); - program->Base.InputsRead = 0x0; - program->Base.OutputsWritten = 0x0; - program->Position = 0; - program->MajorVersion = program->MinorVersion = 0; - program->PrecisionOption = GL_DONT_CARE; - program->FogOption = GL_NONE; - program->HintPositionInvariant = GL_FALSE; - for (a = 0; a < MAX_TEXTURE_IMAGE_UNITS; a++) - program->TexturesUsed[a] = 0x0; - program->ShadowSamplers = 0x0; - program->NumAluInstructions = - program->NumTexInstructions = - program->NumTexIndirections = 0; - program->UsesKill = 0; - - vc_head = NULL; - err = GL_FALSE; - - /* Start examining the tokens in the array */ - inst = parsed; - - /* Check the grammer rev */ - if (*inst++ != REVISION) { - program_error (ctx, 0, "Grammar version mismatch"); - err = GL_TRUE; - } - else { - /* ignore program target */ - inst++; - err = parse_instructions(ctx, inst, &vc_head, program); - } - - /*debug_variables(ctx, vc_head, program); */ - - /* We're done with the parsed binary array */ - var_cache_destroy (&vc_head); - - _mesa_free (parsed); - - /* Reallocate the instruction array from size [MAX_PROGRAM_INSTRUCTIONS] - * to size [ap.Base.NumInstructions]. - */ - program->Base.Instructions - = _mesa_realloc_instructions(program->Base.Instructions, - MAX_PROGRAM_INSTRUCTIONS, - program->Base.NumInstructions); - - return !err; -} - +#include "program_parser.h" void @@ -3932,11 +71,18 @@ _mesa_parse_arb_fragment_program(GLcontext* ctx, GLenum target, const GLvoid *str, GLsizei len, struct gl_fragment_program *program) { - struct arb_program ap; + struct gl_program prog; + struct asm_parser_state state; GLuint i; ASSERT(target == GL_FRAGMENT_PROGRAM_ARB); - if (!_mesa_parse_arb_program(ctx, target, (const GLubyte*) str, len, &ap)) { + + memset(&prog, 0, sizeof(prog)); + memset(&state, 0, sizeof(state)); + state.prog = &prog; + + if (!_mesa_parse_arb_program(ctx, target, (const GLubyte*) str, len, + &state)) { /* Error in the program. Just return. */ return; } @@ -3944,44 +90,50 @@ _mesa_parse_arb_fragment_program(GLcontext* ctx, GLenum target, /* Copy the relevant contents of the arb_program struct into the * fragment_program struct. */ - program->Base.String = ap.Base.String; - program->Base.NumInstructions = ap.Base.NumInstructions; - program->Base.NumTemporaries = ap.Base.NumTemporaries; - program->Base.NumParameters = ap.Base.NumParameters; - program->Base.NumAttributes = ap.Base.NumAttributes; - program->Base.NumAddressRegs = ap.Base.NumAddressRegs; - program->Base.NumNativeInstructions = ap.Base.NumNativeInstructions; - program->Base.NumNativeTemporaries = ap.Base.NumNativeTemporaries; - program->Base.NumNativeParameters = ap.Base.NumNativeParameters; - program->Base.NumNativeAttributes = ap.Base.NumNativeAttributes; - program->Base.NumNativeAddressRegs = ap.Base.NumNativeAddressRegs; - program->Base.NumAluInstructions = ap.Base.NumAluInstructions; - program->Base.NumTexInstructions = ap.Base.NumTexInstructions; - program->Base.NumTexIndirections = ap.Base.NumTexIndirections; - program->Base.NumNativeAluInstructions = ap.Base.NumAluInstructions; - program->Base.NumNativeTexInstructions = ap.Base.NumTexInstructions; - program->Base.NumNativeTexIndirections = ap.Base.NumTexIndirections; - program->Base.InputsRead = ap.Base.InputsRead; - program->Base.OutputsWritten = ap.Base.OutputsWritten; + program->Base.String = prog.String; + program->Base.NumInstructions = prog.NumInstructions; + program->Base.NumTemporaries = prog.NumTemporaries; + program->Base.NumParameters = prog.NumParameters; + program->Base.NumAttributes = prog.NumAttributes; + program->Base.NumAddressRegs = prog.NumAddressRegs; + program->Base.NumNativeInstructions = prog.NumNativeInstructions; + program->Base.NumNativeTemporaries = prog.NumNativeTemporaries; + program->Base.NumNativeParameters = prog.NumNativeParameters; + program->Base.NumNativeAttributes = prog.NumNativeAttributes; + program->Base.NumNativeAddressRegs = prog.NumNativeAddressRegs; + program->Base.NumAluInstructions = prog.NumAluInstructions; + program->Base.NumTexInstructions = prog.NumTexInstructions; + program->Base.NumTexIndirections = prog.NumTexIndirections; + program->Base.NumNativeAluInstructions = prog.NumAluInstructions; + program->Base.NumNativeTexInstructions = prog.NumTexInstructions; + program->Base.NumNativeTexIndirections = prog.NumTexIndirections; + program->Base.InputsRead = prog.InputsRead; + program->Base.OutputsWritten = prog.OutputsWritten; for (i = 0; i < MAX_TEXTURE_IMAGE_UNITS; i++) { - program->Base.TexturesUsed[i] = ap.TexturesUsed[i]; - if (ap.TexturesUsed[i]) + program->Base.TexturesUsed[i] = prog.TexturesUsed[i]; + if (prog.TexturesUsed[i]) program->Base.SamplersUsed |= (1 << i); } - program->Base.ShadowSamplers = ap.ShadowSamplers; - program->FogOption = ap.FogOption; - program->UsesKill = ap.UsesKill; + program->Base.ShadowSamplers = prog.ShadowSamplers; + switch (state.option.Fog) { + case OPTION_FOG_EXP: program->FogOption = GL_EXP; break; + case OPTION_FOG_EXP2: program->FogOption = GL_EXP2; break; + case OPTION_FOG_LINEAR: program->FogOption = GL_LINEAR; break; + default: program->FogOption = GL_NONE; break; + } + + program->UsesKill = state.fragment.UsesKill; if (program->FogOption) program->Base.InputsRead |= FRAG_BIT_FOGC; if (program->Base.Instructions) _mesa_free(program->Base.Instructions); - program->Base.Instructions = ap.Base.Instructions; + program->Base.Instructions = prog.Instructions; if (program->Base.Parameters) _mesa_free_parameter_list(program->Base.Parameters); - program->Base.Parameters = ap.Base.Parameters; + program->Base.Parameters = prog.Parameters; /* Append fog instructions now if the program has "OPTION ARB_fog_exp" * or similar. We used to leave this up to drivers, but it appears @@ -4011,11 +163,17 @@ _mesa_parse_arb_vertex_program(GLcontext *ctx, GLenum target, const GLvoid *str, GLsizei len, struct gl_vertex_program *program) { - struct arb_program ap; + struct gl_program prog; + struct asm_parser_state state; ASSERT(target == GL_VERTEX_PROGRAM_ARB); - if (!_mesa_parse_arb_program(ctx, target, (const GLubyte*) str, len, &ap)) { + memset(&prog, 0, sizeof(prog)); + memset(&state, 0, sizeof(state)); + state.prog = &prog; + + if (!_mesa_parse_arb_program(ctx, target, (const GLubyte*) str, len, + &state)) { _mesa_error(ctx, GL_INVALID_OPERATION, "glProgramString(bad program)"); return; } @@ -4023,28 +181,29 @@ _mesa_parse_arb_vertex_program(GLcontext *ctx, GLenum target, /* Copy the relevant contents of the arb_program struct into the * vertex_program struct. */ - program->Base.String = ap.Base.String; - program->Base.NumInstructions = ap.Base.NumInstructions; - program->Base.NumTemporaries = ap.Base.NumTemporaries; - program->Base.NumParameters = ap.Base.NumParameters; - program->Base.NumAttributes = ap.Base.NumAttributes; - program->Base.NumAddressRegs = ap.Base.NumAddressRegs; - program->Base.NumNativeInstructions = ap.Base.NumNativeInstructions; - program->Base.NumNativeTemporaries = ap.Base.NumNativeTemporaries; - program->Base.NumNativeParameters = ap.Base.NumNativeParameters; - program->Base.NumNativeAttributes = ap.Base.NumNativeAttributes; - program->Base.NumNativeAddressRegs = ap.Base.NumNativeAddressRegs; - program->Base.InputsRead = ap.Base.InputsRead; - program->Base.OutputsWritten = ap.Base.OutputsWritten; - program->IsPositionInvariant = ap.HintPositionInvariant; + program->Base.String = prog.String; + program->Base.NumInstructions = prog.NumInstructions; + program->Base.NumTemporaries = prog.NumTemporaries; + program->Base.NumParameters = prog.NumParameters; + program->Base.NumAttributes = prog.NumAttributes; + program->Base.NumAddressRegs = prog.NumAddressRegs; + program->Base.NumNativeInstructions = prog.NumNativeInstructions; + program->Base.NumNativeTemporaries = prog.NumNativeTemporaries; + program->Base.NumNativeParameters = prog.NumNativeParameters; + program->Base.NumNativeAttributes = prog.NumNativeAttributes; + program->Base.NumNativeAddressRegs = prog.NumNativeAddressRegs; + program->Base.InputsRead = prog.InputsRead; + program->Base.OutputsWritten = prog.OutputsWritten; + program->IsPositionInvariant = (state.option.PositionInvariant) + ? GL_TRUE : GL_FALSE; if (program->Base.Instructions) _mesa_free(program->Base.Instructions); - program->Base.Instructions = ap.Base.Instructions; + program->Base.Instructions = prog.Instructions; if (program->Base.Parameters) _mesa_free_parameter_list(program->Base.Parameters); - program->Base.Parameters = ap.Base.Parameters; + program->Base.Parameters = prog.Parameters; #if DEBUG_VP _mesa_printf("____________Vertex program %u __________\n", program->Base.Id); diff --git a/src/mesa/shader/arbprogram.syn b/src/mesa/shader/arbprogram.syn deleted file mode 100644 index b12c6a0eda5..00000000000 --- a/src/mesa/shader/arbprogram.syn +++ /dev/null @@ -1,2824 +0,0 @@ -/* - * Mesa 3-D graphics library - * Version: 6.2 - * - * Copyright (C) 1999-2004 Brian Paul All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - - /** - * \file arbprogram.syn - * ARB_fragment/vertex_program syntax - * \author Michal Krol - */ - -.syntax program; - -/* - This value must be incremented every time emit code values or structure of the production - array changes. This value is placed at the beginning of the production array. The loader - compares the value with its REVISION value. If they do not match, the loader is not up - to date. -*/ -.emtcode REVISION 0x0a - -/* program type */ -.emtcode FRAGMENT_PROGRAM 0x01 -.emtcode VERTEX_PROGRAM 0x02 - -/* program section */ -.emtcode OPTION 0x01 -.emtcode INSTRUCTION 0x02 -.emtcode DECLARATION 0x03 -.emtcode END 0x04 - -/* GL_ARB_fragment_program option */ -.emtcode ARB_PRECISION_HINT_FASTEST 0x00 -.emtcode ARB_PRECISION_HINT_NICEST 0x01 -.emtcode ARB_FOG_EXP 0x02 -.emtcode ARB_FOG_EXP2 0x03 -.emtcode ARB_FOG_LINEAR 0x04 - -/* GL_ARB_vertex_program option */ -.emtcode ARB_POSITION_INVARIANT 0x05 - -/* GL_ARB_fragment_program_shadow option */ -.emtcode ARB_FRAGMENT_PROGRAM_SHADOW 0x06 - -/* GL_ARB_draw_buffers option */ -.emtcode ARB_DRAW_BUFFERS 0x07 - -/* GL_MESA_texture_array option */ -.emtcode MESA_TEXTURE_ARRAY 0x08 - -/* GL_ARB_fragment_program instruction class */ -.emtcode OP_ALU_INST 0x00 -.emtcode OP_TEX_INST 0x01 - -/* GL_ARB_vertex_program instruction class */ -/* OP_ALU_INST */ - -/* GL_ARB_fragment_program instruction type */ -.emtcode OP_ALU_VECTOR 0x00 -.emtcode OP_ALU_SCALAR 0x01 -.emtcode OP_ALU_BINSC 0x02 -.emtcode OP_ALU_BIN 0x03 -.emtcode OP_ALU_TRI 0x04 -.emtcode OP_ALU_SWZ 0x05 -.emtcode OP_TEX_SAMPLE 0x06 -.emtcode OP_TEX_KIL 0x07 - -/* GL_ARB_vertex_program instruction type */ -.emtcode OP_ALU_ARL 0x08 -/* OP_ALU_VECTOR */ -/* OP_ALU_SCALAR */ -/* OP_ALU_BINSC */ -/* OP_ALU_BIN */ -/* OP_ALU_TRI */ -/* OP_ALU_SWZ */ - -/* GL_ARB_fragment_program instruction code */ -.emtcode OP_ABS 0x00 -.emtcode OP_ABS_SAT 0x1B -.emtcode OP_FLR 0x09 -.emtcode OP_FLR_SAT 0x26 -.emtcode OP_FRC 0x0A -.emtcode OP_FRC_SAT 0x27 -.emtcode OP_LIT 0x0C -.emtcode OP_LIT_SAT 0x2A -.emtcode OP_MOV 0x11 -.emtcode OP_MOV_SAT 0x30 -.emtcode OP_COS 0x1F -.emtcode OP_COS_SAT 0x20 -.emtcode OP_EX2 0x07 -.emtcode OP_EX2_SAT 0x25 -.emtcode OP_LG2 0x0B -.emtcode OP_LG2_SAT 0x29 -.emtcode OP_RCP 0x14 -.emtcode OP_RCP_SAT 0x33 -.emtcode OP_RSQ 0x15 -.emtcode OP_RSQ_SAT 0x34 -.emtcode OP_SIN 0x38 -.emtcode OP_SIN_SAT 0x39 -.emtcode OP_SCS 0x35 -.emtcode OP_SCS_SAT 0x36 -.emtcode OP_POW 0x13 -.emtcode OP_POW_SAT 0x32 -.emtcode OP_ADD 0x01 -.emtcode OP_ADD_SAT 0x1C -.emtcode OP_DP3 0x03 -.emtcode OP_DP3_SAT 0x21 -.emtcode OP_DP4 0x04 -.emtcode OP_DP4_SAT 0x22 -.emtcode OP_DPH 0x05 -.emtcode OP_DPH_SAT 0x23 -.emtcode OP_DST 0x06 -.emtcode OP_DST_SAT 0x24 -.emtcode OP_MAX 0x0F -.emtcode OP_MAX_SAT 0x2E -.emtcode OP_MIN 0x10 -.emtcode OP_MIN_SAT 0x2F -.emtcode OP_MUL 0x12 -.emtcode OP_MUL_SAT 0x31 -.emtcode OP_SGE 0x16 -.emtcode OP_SGE_SAT 0x37 -.emtcode OP_SLT 0x17 -.emtcode OP_SLT_SAT 0x3A -.emtcode OP_SUB 0x18 -.emtcode OP_SUB_SAT 0x3B -.emtcode OP_XPD 0x1A -.emtcode OP_XPD_SAT 0x43 -.emtcode OP_CMP 0x1D -.emtcode OP_CMP_SAT 0x1E -.emtcode OP_LRP 0x2B -.emtcode OP_LRP_SAT 0x2C -.emtcode OP_MAD 0x0E -.emtcode OP_MAD_SAT 0x2D -.emtcode OP_SWZ 0x19 -.emtcode OP_SWZ_SAT 0x3C -.emtcode OP_TEX 0x3D -.emtcode OP_TEX_SAT 0x3E -.emtcode OP_TXB 0x3F -.emtcode OP_TXB_SAT 0x40 -.emtcode OP_TXP 0x41 -.emtcode OP_TXP_SAT 0x42 -.emtcode OP_KIL 0x28 - -/* GL_ARB_vertex_program instruction code */ -.emtcode OP_ARL 0x02 -/* OP_ABS */ -/* OP_FLR */ -/* OP_FRC */ -/* OP_LIT */ -/* OP_MOV */ -/* OP_EX2 */ -.emtcode OP_EXP 0x08 -/* OP_LG2 */ -.emtcode OP_LOG 0x0D -/* OP_RCP */ -/* OP_RSQ */ -/* OP_POW */ -/* OP_ADD */ -/* OP_DP3 */ -/* OP_DP4 */ -/* OP_DPH */ -/* OP_DST */ -/* OP_MAX */ -/* OP_MIN */ -/* OP_MUL */ -/* OP_SGE */ -/* OP_SLT */ -/* OP_SUB */ -/* OP_XPD */ -/* OP_MAD */ -/* OP_SWZ */ - -/* fragment attribute binding */ -.emtcode FRAGMENT_ATTRIB_COLOR 0x01 -.emtcode FRAGMENT_ATTRIB_TEXCOORD 0x02 -.emtcode FRAGMENT_ATTRIB_FOGCOORD 0x03 -.emtcode FRAGMENT_ATTRIB_POSITION 0x04 - -/* vertex attribute binding */ -.emtcode VERTEX_ATTRIB_POSITION 0x01 -.emtcode VERTEX_ATTRIB_WEIGHT 0x02 -.emtcode VERTEX_ATTRIB_NORMAL 0x03 -.emtcode VERTEX_ATTRIB_COLOR 0x04 -.emtcode VERTEX_ATTRIB_FOGCOORD 0x05 -.emtcode VERTEX_ATTRIB_TEXCOORD 0x06 -.emtcode VERTEX_ATTRIB_MATRIXINDEX 0x07 -.emtcode VERTEX_ATTRIB_GENERIC 0x08 - -/* fragment result binding */ -.emtcode FRAGMENT_RESULT_COLOR 0x01 -.emtcode FRAGMENT_RESULT_DEPTH 0x02 - -/* vertex result binding */ -.emtcode VERTEX_RESULT_POSITION 0x01 -.emtcode VERTEX_RESULT_COLOR 0x02 -.emtcode VERTEX_RESULT_FOGCOORD 0x03 -.emtcode VERTEX_RESULT_POINTSIZE 0x04 -.emtcode VERTEX_RESULT_TEXCOORD 0x05 - -/* texture target */ -.emtcode TEXTARGET_1D 0x01 -.emtcode TEXTARGET_2D 0x02 -.emtcode TEXTARGET_3D 0x03 -.emtcode TEXTARGET_RECT 0x04 -.emtcode TEXTARGET_CUBE 0x05 -/* GL_ARB_fragment_program_shadow */ -.emtcode TEXTARGET_SHADOW1D 0x06 -.emtcode TEXTARGET_SHADOW2D 0x07 -.emtcode TEXTARGET_SHADOWRECT 0x08 -/* GL_MESA_texture_array */ -.emtcode TEXTARGET_1D_ARRAY 0x09 -.emtcode TEXTARGET_2D_ARRAY 0x0a -.emtcode TEXTARGET_SHADOW1D_ARRAY 0x0b -.emtcode TEXTARGET_SHADOW2D_ARRAY 0x0c - -/* face type */ -.emtcode FACE_FRONT 0x00 -.emtcode FACE_BACK 0x01 - -/* color type */ -.emtcode COLOR_PRIMARY 0x00 -.emtcode COLOR_SECONDARY 0x01 - -/* component */ -.emtcode COMPONENT_X 0x00 -.emtcode COMPONENT_Y 0x01 -.emtcode COMPONENT_Z 0x02 -.emtcode COMPONENT_W 0x03 -.emtcode COMPONENT_0 0x04 -.emtcode COMPONENT_1 0x05 - -/* array index type */ -.emtcode ARRAY_INDEX_ABSOLUTE 0x00 -.emtcode ARRAY_INDEX_RELATIVE 0x01 - -/* matrix name */ -.emtcode MATRIX_MODELVIEW 0x01 -.emtcode MATRIX_PROJECTION 0x02 -.emtcode MATRIX_MVP 0x03 -.emtcode MATRIX_TEXTURE 0x04 -.emtcode MATRIX_PALETTE 0x05 -.emtcode MATRIX_PROGRAM 0x06 - -/* matrix modifier */ -.emtcode MATRIX_MODIFIER_IDENTITY 0x00 -.emtcode MATRIX_MODIFIER_INVERSE 0x01 -.emtcode MATRIX_MODIFIER_TRANSPOSE 0x02 -.emtcode MATRIX_MODIFIER_INVTRANS 0x03 - -/* constant type */ -.emtcode CONSTANT_SCALAR 0x01 -.emtcode CONSTANT_VECTOR 0x02 - -/* program param type */ -.emtcode PROGRAM_PARAM_ENV 0x01 -.emtcode PROGRAM_PARAM_LOCAL 0x02 - -/* register type */ -.emtcode REGISTER_ATTRIB 0x01 -.emtcode REGISTER_PARAM 0x02 -.emtcode REGISTER_RESULT 0x03 -.emtcode REGISTER_ESTABLISHED_NAME 0x04 - -/* param binding */ -.emtcode PARAM_NULL 0x00 -.emtcode PARAM_ARRAY_ELEMENT 0x01 -.emtcode PARAM_STATE_ELEMENT 0x02 -.emtcode PARAM_PROGRAM_ELEMENT 0x03 -.emtcode PARAM_PROGRAM_ELEMENTS 0x04 -.emtcode PARAM_CONSTANT 0x05 - -/* param state property */ -.emtcode STATE_MATERIAL 0x01 -.emtcode STATE_LIGHT 0x02 -.emtcode STATE_LIGHT_MODEL 0x03 -.emtcode STATE_LIGHT_PROD 0x04 -.emtcode STATE_FOG 0x05 -.emtcode STATE_MATRIX_ROWS 0x06 -/* GL_ARB_fragment_program */ -.emtcode STATE_TEX_ENV 0x07 -.emtcode STATE_DEPTH 0x08 -/* GL_ARB_vertex_program */ -.emtcode STATE_TEX_GEN 0x09 -.emtcode STATE_CLIP_PLANE 0x0A -.emtcode STATE_POINT 0x0B - -/* state material property */ -.emtcode MATERIAL_AMBIENT 0x01 -.emtcode MATERIAL_DIFFUSE 0x02 -.emtcode MATERIAL_SPECULAR 0x03 -.emtcode MATERIAL_EMISSION 0x04 -.emtcode MATERIAL_SHININESS 0x05 - -/* state light property */ -.emtcode LIGHT_AMBIENT 0x01 -.emtcode LIGHT_DIFFUSE 0x02 -.emtcode LIGHT_SPECULAR 0x03 -.emtcode LIGHT_POSITION 0x04 -.emtcode LIGHT_ATTENUATION 0x05 -.emtcode LIGHT_HALF 0x06 -.emtcode LIGHT_SPOT_DIRECTION 0x07 - -/* state light model property */ -.emtcode LIGHT_MODEL_AMBIENT 0x01 -.emtcode LIGHT_MODEL_SCENECOLOR 0x02 - -/* state light product property */ -.emtcode LIGHT_PROD_AMBIENT 0x01 -.emtcode LIGHT_PROD_DIFFUSE 0x02 -.emtcode LIGHT_PROD_SPECULAR 0x03 - -/* state texture environment property */ -.emtcode TEX_ENV_COLOR 0x01 - -/* state texture generation coord property */ -.emtcode TEX_GEN_EYE 0x01 -.emtcode TEX_GEN_OBJECT 0x02 - -/* state fog property */ -.emtcode FOG_COLOR 0x01 -.emtcode FOG_PARAMS 0x02 - -/* state depth property */ -.emtcode DEPTH_RANGE 0x01 - -/* state point parameters property */ -.emtcode POINT_SIZE 0x01 -.emtcode POINT_ATTENUATION 0x02 - -/* declaration */ -.emtcode ATTRIB 0x01 -.emtcode PARAM 0x02 -.emtcode TEMP 0x03 -.emtcode OUTPUT 0x04 -.emtcode ALIAS 0x05 -/* GL_ARB_vertex_program */ -.emtcode ADDRESS 0x06 - -/* error messages */ -.errtext UNKNOWN_PROGRAM_SIGNATURE "1001: '$e_signature$': unknown program signature" -.errtext MISSING_END_OR_INVALID_STATEMENT "1002: '$e_statement$': invalid statement" -.errtext CODE_AFTER_END "1003: '$e_statement$': code after 'END' keyword" -.errtext INVALID_PROGRAM_OPTION "1004: '$e_identifier$': invalid program option" -.errtext EXT_SWIZ_COMP_EXPECTED "1005: extended swizzle component expected but '$e_token$' found" -.errtext TEX_TARGET_EXPECTED "1006: texture target expected but '$e_token$' found" -.errtext TEXTURE_EXPECTED "1007: 'texture' expected but '$e_identifier$' found" -.errtext SOURCE_REGISTER_EXPECTED "1008: source register expected but '$e_token$' found" -.errtext DESTINATION_REGISTER_EXPECTED "1009: destination register expected but '$e_token$' found" -.errtext INVALID_ADDRESS_COMPONENT "1010: '$e_identifier$': invalid address component" -.errtext INVALID_ADDRESS_WRITEMASK "1011: '$e_identifier$': invalid address writemask" -.errtext INVALID_COMPONENT "1012: '$e_charordigit$': invalid component" -.errtext INVALID_SUFFIX "1013: '$e_identifier$': invalid suffix" -.errtext INVALID_WRITEMASK "1014: '$e_identifier$': invalid writemask" -.errtext FRAGMENT_EXPECTED "1015: 'fragment' expected but '$e_identifier$' found" -.errtext VERTEX_EXPECTED "1016: 'vertex' expected but '$e_identifier$' found" -.errtext INVALID_FRAGMENT_PROPERTY "1017: '$e_identifier$': invalid fragment property" -.errtext INVALID_VERTEX_PROPERTY "1018: '$e_identifier$': invalid vertex property" -.errtext INVALID_STATE_PROPERTY "1019: '$e_identifier$': invalid state property" -.errtext INVALID_MATERIAL_PROPERTY "1020: '$e_identifier$': invalid material property" -.errtext INVALID_LIGHT_PROPERTY "1021: '$e_identifier$': invalid light property" -.errtext INVALID_SPOT_PROPERTY "1022: '$e_identifier$': invalid spot property" -.errtext INVALID_LIGHTMODEL_PROPERTY "1023: '$e_identifier$': invalid light model property" -.errtext INVALID_LIGHTPROD_PROPERTY "1024: '$e_identifier$': invalid light product property" -.errtext INVALID_TEXENV_PROPERTY "1025: '$e_identifier$': invalid texture environment property" -.errtext INVALID_TEXGEN_PROPERTY "1026: '$e_identifier$': invalid texture generating property" -.errtext INVALID_TEXGEN_COORD "1027: '$e_identifier$': invalid texture generating coord" -.errtext INVALID_FOG_PROPERTY "1028: '$e_identifier$': invalid fog property" -.errtext INVALID_DEPTH_PROPERTY "1029: '$e_identifier$': invalid depth property" -.errtext INVALID_CLIPPLANE_PROPERTY "1030: '$e_identifier$': invalid clip plane property" -.errtext INVALID_POINT_PROPERTY "1031: '$e_identifier$': invalid point property" -.errtext MATRIX_ROW_SELECTOR_OR_MODIFIER_EXPECTED "1032: matrix row selector or modifier expected but '$e_token$' found" -.errtext INVALID_MATRIX_NAME "1033: '$e_identifier$': invalid matrix name" -.errtext INVALID_PROGRAM_PROPERTY "1034: '$e_identifier$': invalid program property" -.errtext RESULT_EXPECTED "1035: 'result' expected but '$e_token$' found" -.errtext INVALID_RESULT_PROPERTY "1036: '$e_identifier$': invalid result property" -.errtext INVALID_FACE_PROPERTY "1037: '$e_identifier$': invalid face property" -.errtext INVALID_COLOR_PROPERTY "1038: '$e_identifier$': invalid color property" -.errtext IDENTIFIER_EXPECTED "1039: identifier expected but '$e_token$' found" -.errtext RESERVED_KEYWORD "1040: use of reserved keyword as an identifier" -.errtext INTEGER_EXPECTED "1041: integer value expected but '$e_token$' found" -.errtext MISSING_SEMICOLON "1042: ';' expected but '$e_token$' found" -.errtext MISSING_COMMA "1043: ',' expected but '$e_token$' found" -.errtext MISSING_LBRACKET "1044: '[' expected but '$e_token$' found" -.errtext MISSING_RBRACKET "1045: ']' expected but '$e_token$' found" -.errtext MISSING_DOT "1046: '.' expected but '$e_token$' found" -.errtext MISSING_EQUAL "1047: '=' expected but '$e_token$' found" -.errtext MISSING_LBRACE "1048: '{' expected but '$e_token$' found" -.errtext MISSING_RBRACE "1049: '}' expected but '$e_token$' found" -.errtext MISSING_DOTDOT "1050: '..' expected but '$e_token$' found" -.errtext MISSING_FRACTION_OR_EXPONENT "1051: missing fraction part or exponent" -.errtext MISSING_DOT_OR_EXPONENT "1052: missing '.' or exponent" -.errtext EXPONENT_VALUE_EXPECTED "1053: exponent value expected" -.errtext INTEGER_OUT_OF_RANGE "1054: integer value out of range" -.errtext OPERATION_NEEDS_DESTINATION_VARIABLE "1055: operation needs destination variable" -.errtext OPERATION_NEEDS_SOURCE_VARIABLE "1056: operation needs source variable" -.errtext ADDRESS_REGISTER_EXPECTED "1057: address register expected but '$e_token$' found" -.errtext ADDRESS_REGISTER_OR_INTEGER_EXPECTED "1058: address register or integer literal expected but '$e_token$' found" - -/* extension presence condition registers */ - -/* GL_ARB_vertex_blend */ -/* GL_EXT_vertex_weighting */ -.regbyte vertex_blend 0x00 - -/* GL_ARB_matrix_palette */ -.regbyte matrix_palette 0x00 - -/* GL_ARB_point_parameters */ -/* GL_EXT_point_parameters */ -.regbyte point_parameters 0x00 - -/* GL_EXT_secondary_color */ -.regbyte secondary_color 0x00 - -/* GL_EXT_fog_coord */ -.regbyte fog_coord 0x00 - -/* GL_EXT_texture_rectangle */ -/* GL_NV_texture_rectangle */ -.regbyte texture_rectangle 0x00 - -/* GL_ARB_fragment_program_shadow */ -.regbyte fragment_program_shadow 0x00 - -/* GL_ARB_draw_buffers */ -.regbyte draw_buffers 0x00 - -/* GL_MESA_texture_array */ -.regbyte texture_array 0x00 - -/* option presence condition registers */ -/* they are all initially set to zero - when a particular OPTION is encountered, the appropriate */ -/* register is set to 1 to indicate that the OPTION was specified. */ - -/* GL_ARB_fragment_program */ -.regbyte ARB_precision_hint_fastest 0x00 -.regbyte ARB_precision_hint_nicest 0x00 -.regbyte ARB_fog_exp 0x00 -.regbyte ARB_fog_exp2 0x00 -.regbyte ARB_fog_linear 0x00 - -/* GL_ARB_vertex_program */ -.regbyte ARB_position_invariant 0x00 - -/* GL_ARB_fragment_program_shadow */ -.regbyte ARB_fragment_program_shadow 0x00 - -/* GL_ARB_draw_buffers */ -.regbyte ARB_draw_buffers 0x00 - -/* GL_MESA_texture_array */ -.regbyte MESA_texture_array 0x00 - -/* program target condition register */ -/* this syntax script deals with two program targets - VERTEX_PROGRAM and FRAGMENT_PROGRAM. */ -/* to distinguish between them we need a register that will store for us the current target. */ -/* the client will typically set the register to apropriate value before parsing a particular */ -/* program. the mapping between program targets and their values is listed below. */ -/* */ -/* program target register value */ -/* ---------------------------------------------- */ -/* FRAGMENT_PROGRAM 0x10 */ -/* VERTEX_PROGRAM 0x20 */ -/* */ -/* the initial value of the register is 0 to catch potential errors with not setting the register */ -/* with the proper value. */ -.regbyte program_target 0x00 - -/* - ::= "END" -*/ -program - programs .error UNKNOWN_PROGRAM_SIGNATURE .emit REVISION; -programs - .if (program_target == 0x10) frag_program_1_0 .emit FRAGMENT_PROGRAM .emit 0x01 .emit 0x00 .or - .if (program_target == 0x20) vert_program_1_0 .emit VERTEX_PROGRAM .emit 0x01 .emit 0x00; -frag_program_1_0 - '!' .and '!' .and 'A' .and 'R' .and 'B' .and 'f' .and 'p' .and '1' .and '.' .and '0' .and - optional_space .and fp_optionSequence .and fp_statementSequence .and - "END" .error MISSING_END_OR_INVALID_STATEMENT .emit END .and optional_space .and - '\0' .error CODE_AFTER_END; -vert_program_1_0 - '!' .and '!' .and 'A' .and 'R' .and 'B' .and 'v' .and 'p' .and '1' .and '.' .and '0' .and - optional_space .and vp_optionSequence .and vp_statementSequence .and - "END" .error MISSING_END_OR_INVALID_STATEMENT .emit END .and optional_space .and - '\0' .error CODE_AFTER_END; - -/* - ::=